US9183493B2 - Adaptive plasticity apparatus and methods for spiking neuron network - Google Patents

Adaptive plasticity apparatus and methods for spiking neuron network Download PDF

Info

Publication number
US9183493B2
US9183493B2 US13/660,923 US201213660923A US9183493B2 US 9183493 B2 US9183493 B2 US 9183493B2 US 201213660923 A US201213660923 A US 201213660923A US 9183493 B2 US9183493 B2 US 9183493B2
Authority
US
United States
Prior art keywords
connection
neuron
input
synaptic
time
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active, expires
Application number
US13/660,923
Other versions
US20140122397A1 (en
Inventor
Micah Richert
Botond Szatmary
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Brain Corp
Original Assignee
Brain Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Brain Corp filed Critical Brain Corp
Priority to US13/660,923 priority Critical patent/US9183493B2/en
Assigned to BRAIN CORPORATION reassignment BRAIN CORPORATION ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: SZATMARY, BOTOND, RICHERT, MICAH
Publication of US20140122397A1 publication Critical patent/US20140122397A1/en
Application granted granted Critical
Publication of US9183493B2 publication Critical patent/US9183493B2/en
Assigned to HERCULES CAPITAL, INC. reassignment HERCULES CAPITAL, INC. SECURITY INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: BRAIN CORPORATION
Active legal-status Critical Current
Adjusted expiration legal-status Critical

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/049Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/06Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
    • G06N3/063Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using electronic means
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • G06N3/088Non-supervised learning, e.g. competitive learning

Definitions

  • the present disclosure relates generally to artificial neural networks, and more particularly in one exemplary aspect to computer apparatus and methods for plasticity implementation in a pulse-code neural network.
  • artificial spiking neural networks may comprise a plurality of units (or nodes), which correspond to neurons in a biological neural network. Any given unit may be connected to many other units via connections, also referred to as communications channels, and/or synaptic connections.
  • the units providing inputs to any given unit are commonly referred to as the pre-synaptic units, while the unit receiving the inputs is referred to as the post-synaptic unit.
  • Each of the unit-to-unit connections may be assigned, inter alia, a connection efficacy, which in general may refer to a magnitude and/or probability of input spike influence on unit output response (i.e., output spike generation/firing).
  • the efficacy may comprise, for example a parameter—synaptic weight—by which one or more state variables of post-synaptic unit are changed.
  • synaptic weights may be dynamically adjusted using what is referred to as the spike-timing dependent plasticity (STDP) in order to implement, among other things, network learning.
  • STDP spike-timing dependent plasticity
  • larger weights may be associated with a greater effect a synapse has on the activity of the post-synaptic neuron.
  • connections that deliver inputs (to a given unit) prior to generation of post-synaptic response may be potentiated, while connections that deliver inputs after the generation of the post-synaptic response may be depressed.
  • the choice of plasticity functional dependence may determine network behavior.
  • various implementations plasticity mechanisms exist including, for example, the use of target connection efficiency (that may be defined as a ratio of a number of input (pre-synaptic) spikes N fire delivered to a neuron via the connection that are followed by neuron response (e.g., post-synaptic spike) generation, to the total number of input spikes N tot delivered to the neuron via the connection.
  • target connection efficiency that may be defined as a ratio of a number of input (pre-synaptic) spikes N fire delivered to a neuron via the connection that are followed by neuron response (e.g., post-synaptic spike) generation, to the total number of input spikes N tot delivered to the neuron via the connection
  • the present disclosure satisfies the foregoing needs by providing, inter alia, apparatus and methods for implementing adaptive plasticity in spiking neuron networks that may be dynamically adjusted in accordance with the connection activity thereby enhancing the learning performance of a neural network.
  • a method of updating an efficacy of a connection configured to communicate an input to an artificial spiking neuron includes: determining a similarity measure between the input and a neuron output; identifying one or more components associated with one or more lag values, respectively, from the similarity measure; and updating the efficacy based on the one or more components.
  • the apparatus comprise a storage medium, the storage medium comprising a plurality of executable instructions configured to, when executed, adjust an efficacy of an interface of the neuron by at least: adjustment of neuron excitability based on one or more pre-synaptic inputs; storage of one or more pre-synaptic times associated with one or more pre-synaptic inputs; and when the excitability is above a threshold: generation of post-synaptic output; determination of a similarity measure between the neuron output signal, comprising the response, and the one or more pre-synaptic inputs; determination of a long term potentiation component (LTP) of the adaptive plasticity based on the similarity measure; and determination of a long term depression component of the adaptive plasticity based on a parameter associated with the LTP
  • LTP long term potentiation component
  • the response is generated by the neuron based on the one or more pre-synaptic inputs.
  • a computerized spiking neuron network system configured to implement an adaptive plasticity in the network, and includes: one or more processors configured to execute computer program modules, the execution of the computer program modules configured to cause the one or more processors, based on a response by a neuron of the network, to adjust one or more connections capable of providing input to the neuron by at least: determination of one or more similarity measures between the neuron output signal, comprising the response, and one or more input signals provided by the one or more connections, respectively; determination of a long term potentiation component (LTP) of the adaptive plasticity based on an average of the one or more similarity measures; and determination of a long term depression component of the adaptive plasticity based on a parameter associated with the LTP.
  • LTP long term potentiation component
  • computer readable apparatus in another aspect, includes at least one computer program configured to, when executed, implement an adaptive plasticity in a neural network.
  • FIG. 1 is a block diagram depicting an artificial spiking neural network according to one implementation.
  • FIG. 2 is a graphical illustration depicting histogram determination based on post-synaptic and pre-synaptic activity of a unit of the spiking network of FIG. 1 , according to one implementation.
  • FIG. 3 is a plot depicting adaptively constructed spike-time dependent plasticity (STDP) for use in the spiking network of FIG. 1 , according to one implementation.
  • STDP spike-time dependent plasticity
  • FIG. 4 is a logical flow diagram illustrating determination of an adaptive plasticity mechanism, in accordance with one implementation.
  • FIG. 5 is a logical flow diagram illustrating a method of determining adaptive plasticity for a spiking neuron comprising multiple input connections, in accordance with one implementation.
  • FIG. 6 is a logical flow diagram illustrating a method of connection plasticity update based on the adaptive STDP rule, in accordance with one implementation.
  • FIG. 7 is a block diagram illustrating performance of a sensory processing apparatus comprising adaptive plasticity mechanism in accordance with one implementation.
  • FIG. 8 is a block diagram illustrating a sensory processing apparatus comprising adaptive plasticity mechanism in accordance with one implementation.
  • FIG. 9 is a block diagram illustrating a computerized system useful for, inter alia, providing an adaptive plasticity mechanism in a spiking network, in accordance with one implementation.
  • FIG. 10 is a block diagram illustrating a neuromorphic computerized system useful with, inter alia, adaptive plasticity mechanism in a spiking network, in accordance with one implementation.
  • FIG. 11A is a block diagram illustrating a hierarchical neuromorphic computerized system architecture useful with, inter alia, adaptive plasticity mechanism in a spiking network, in accordance with one implementation.
  • FIG. 11B is a block diagram illustrating a cell-type neuromorphic computerized system architecture useful with, inter alia, adaptive plasticity mechanism in a spiking network, in accordance with one implementation.
  • bus is meant generally to denote all types of interconnection or communication architecture that is used to access the synaptic and neuron memory.
  • the “bus” could be optical, wireless, infrared or another type of communication medium.
  • the exact topology of the bus could be for example standard “bus”, hierarchical bus, network-on-chip, address-event-representation (AER) connection, or other type of communication topology used for accessing, e.g., different memories in pulse-based system.
  • the terms “computer”, “computing device”, and “computerized device”, include, but are not limited to, personal computers (PCs) and minicomputers, whether desktop, laptop, or otherwise, mainframe computers, workstations, servers, personal digital assistants (PDAs), handheld computers, embedded computers, programmable logic device, personal communicators, tablet computers, portable navigation aids, J2ME equipped devices, cellular telephones, smart phones, personal integrated communication or entertainment devices, or literally any other device capable of executing a set of instructions and processing an incoming data signal.
  • PCs personal computers
  • PDAs personal digital assistants
  • handheld computers handheld computers
  • embedded computers embedded computers
  • programmable logic device personal communicators
  • tablet computers tablet computers
  • portable navigation aids J2ME equipped devices
  • J2ME equipped devices J2ME equipped devices
  • cellular telephones cellular telephones
  • smart phones personal integrated communication or entertainment devices
  • personal integrated communication or entertainment devices personal integrated communication or entertainment devices
  • As used herein, the term “computer program” or “software” is meant to include any sequence or human or machine cognizable steps which perform a function.
  • Such program may be rendered in virtually any programming language or environment including, for example, C/C++, C#, Fortran, COBOL, MATLABTM, PASCAL, Python, assembly language, markup languages (e.g., HTML, SGML, XML, VoXML), and the like, as well as object-oriented environments such as the Common Object Request Broker Architecture (CORBA), JavaTM (including J2ME, Java Beans, etc.), Binary Runtime Environment (e.g., BREW), and the like.
  • CORBA Common Object Request Broker Architecture
  • JavaTM including J2ME, Java Beans, etc.
  • BREW Binary Runtime Environment
  • connection As used herein, the terms “connection”, “link”, “synaptic channel”, “transmission channel”, “delay line”, are meant generally to denote a causal link between any two or more entities (whether physical or logical/virtual), which enables information exchange between the entities.
  • memory includes any type of integrated circuit or other storage device adapted for storing digital data including, without limitation, ROM. PROM, EEPROM, DRAM, Mobile DRAM, SDRAM, DDR/2 SDRAM, EDO/FPMS, RLDRAM, SRAM, “flash” memory (e.g., NAND/NOR), memristor memory, and PSRAM.
  • flash memory e.g., NAND/NOR
  • memristor memory and PSRAM.
  • processors are meant generally to include all types of digital processing devices including, without limitation, digital signal processors (DSPs), reduced instruction set computers (RISC), general-purpose (CISC) processors, microprocessors, gate arrays (e.g., field programmable gate arrays (FPGAs)), PLDs, reconfigurable computer fabrics (RCFs), array processors, secure microprocessors, and application-specific integrated circuits (ASICs).
  • DSPs digital signal processors
  • RISC reduced instruction set computers
  • CISC general-purpose processors
  • microprocessors e.g., field programmable gate arrays (FPGAs)), PLDs, reconfigurable computer fabrics (RCFs), array processors, secure microprocessors, and application-specific integrated circuits (ASICs).
  • FPGAs field programmable gate arrays
  • RCFs reconfigurable computer fabrics
  • ASICs application-specific integrated circuits
  • the term “network interface” refers to any signal, data, or software interface with a component, network or process including, without limitation, those of the FireWire (e.g., FW400, FW800, etc.), USB (e.g., USB2), Ethernet (e.g., 10/100, 10/100/1000 (Gigabit Ethernet), 10-Gig-E, etc.), MoCA, Coaxsys (e.g., TVnetTM), radio frequency tuner (e.g., in-band or OOB, cable modem, etc.), Wi-Fi (802.11), WiMAX (802.16), PAN (e.g., 802.15), cellular (e.g., 3G, LTE/LTE-A/TD-LTE, GSM, etc.) or IrDA families.
  • FireWire e.g., FW400, FW800, etc.
  • USB e.g., USB2
  • Ethernet e.g., 10/100, 10/100/1000 (Gigabit Ethernet
  • pulse As used herein, the terms “pulse”, “spike”, “burst of spikes”, and “pulse train” are meant generally to refer to, without limitation, any type of a pulsed signal, e.g., a rapid change in some characteristic of a signal, e.g., amplitude, intensity, phase or frequency, from a baseline value to a higher or lower value, followed by a rapid return to the baseline value and may refer to any of a single spike, a burst of spikes, an electronic pulse, a pulse in voltage, a pulse in electrical current, a software representation of a pulse and/or burst of pulses, a software message representing a discrete pulsed event, and any other pulse or pulse type associated with a discrete information transmission system or mechanism.
  • a pulsed signal e.g., a rapid change in some characteristic of a signal, e.g., amplitude, intensity, phase or frequency, from a baseline value to a higher or lower value, followed by a rapid
  • the term “receptive field” is used to describe sets of weighted inputs from filtered input elements, where the weights may be adjusted.
  • Wi-Fi refers to, without limitation, any of the variants of IEEE-Std. 802.11 or related standards including 802.11 a/b/g/n/s/v and 802.11-2012.
  • wireless means any wireless signal, data, communication, or other interface including without limitation Wi-Fi, Bluetooth, 3G (3GPP/3GPP2), HSDPA/HSUPA, TDMA, CDMA (e.g., IS-95A, WCDMA, etc.), FHSS, DSSS, GSM, PAN/802.15, WiMAX (802.16), 802.20, narrowband/FDMA, OFDM, PCS/DCS, LTE/LTE-A/TD-LTE, analog cellular, CDPD, RFID or NFC (e.g., EPC Global Gen. 2, ISO 14443, ISO 18000-3), satellite systems, millimeter wave or microwave systems, acoustic, and infrared (e.g., IrDA).
  • the present disclosure provides, in one salient aspect, apparatus and methods for implementing adaptive plasticity mechanisms configured to, inter alia, to improve learning (faster learning or learn more useful features) in artificial spiking neuron networks, without requiring pre-defined and/or hard coded plasticity rules.
  • the adaptive plasticity mechanism may be based on a similarity measure (e.g., a cross-correlogram, cross-correlation, convolution, deconvolution, and/or mutual information) between neuron output and neuron input.
  • a similarity measure e.g., a cross-correlogram, cross-correlation, convolution, deconvolution, and/or mutual information
  • the cross-correlogram may be determined based on (i) a time record of pre-synaptic input into the neuron with a time interval ⁇ T prior to the output; and (ii) a time record of post-synaptic output by the neuron with the same time interval.
  • individual correlograms may be constructed for individual connections.
  • An averaged (over multiple connections) correlogram may be determined.
  • multiple correlogram estimates (associates with multiple post-synaptic responses) may be averaged to produce a time-averaged similarity measure.
  • the averaging may be implemented in only time with individual synapses having their respective correlogram. In some implementations, the averaging may be implemented over both time and synapses so that a single correlogram may be shared by all input synapses.
  • individual spikes may be assigned the same amplitude (e.g., binary 1). Accordingly, the binary correlogram may be interpreted as a histogram of pre-synaptic spike occurrence within individual time slots (bins) prior to the post-synaptic response.
  • the STDP rule may comprise for instance a long term depression (LTD) rule and/or a long term potentiation (LTP) rule.
  • LTD long term depression
  • LTP long term potentiation
  • the causal portion (i.e., the portion of the similarity measure for which the pre-synaptic spikes occur before the post spikes) of the similarity measure may be used, inter alia, to construct the LTP rule.
  • Magnitude of plasticity adjustments may be scaled such that the mean of the similarity measure is a constant.
  • magnitude of plasticity adjustments may be scaled such that the maximum value of the similarity measure is a constant.
  • the LTD portion of the STDP rule may comprise a monotonic function (e.g., a constant, gradually increasing and/or decreasing). Magnitude of the LTD portion may be determined based on a statistical parameter of the LTP portion. In some implementations, the statistical parameter may comprise maximum, mean, and/or median of the correlogram).
  • the adaptively configured STDP rules may be used to adjust weights of neuron input connections. In some implementations, the adjustment may be performed based on the post-synaptic response by the neuron.
  • the plasticity methodology describe herein may advantageously improve learning of the neural network.
  • the improved learning may be characterized by a faster convergence, convergence to more meaningful features, fewer recognition errors, requiring fewer neurons to represent a given input.
  • the activity bases plasticity mechanism may enable learning of temporally stable patterns (e.g., learning a component of the temporal feature without drift).
  • adaptive adjustment methodologies are used to implement processing of visual sensory information and feature/object recognition using spiking neuronal networks.
  • Portions of the object recognition apparatus can be embodied for example in a remote computerized apparatus (e.g., server), comprising a computer readable apparatus.
  • Embodiments of the foregoing plasticity mechanism of the present disclosure are useful in a variety of applications including for instance a prosthetic device, autonomous robotic apparatus, and other electromechanical devices requiring visual or other sensory data processing functionality.
  • Implementations of the disclosure may be for example deployed in a hardware and/or software realization of a neuromorphic computer system.
  • a robotic system may include a processor embodied in an application specific integrated circuit, which can be adapted or configured for use in an embedded application (such as a prosthetic device).
  • FIG. 1 illustrates one exemplary implementation of a spiking neuronal network of the disclosure, configured to process sensory information using adaptive plasticity mechanism.
  • the network 100 may comprise one or more spiking neurons, e.g., the neuron 110 in FIG. 1 ),
  • the neuron 110 may be configured to receive feed-forward spiking input via connections 104 .
  • the neuron 110 may be configured to receive feedback spiking input via the connections 124 .
  • the neuron 110 may generate output (e.g., a post-synaptic spike) using any of applicable methodologies such as for example those described in co-owned and co-pending U.S. patent application Ser. No. 13/152,105 filed on Jun.
  • the output spikes of the neuron 110 may be propagated via the connection 114 .
  • Post-synaptic spike generation is well-established in the spiking network arts, and accordingly will not be described in detail herein for brevity and clarity of presentation of the inventive aspects of the present disclosure.
  • connection efficacy may refer to a magnitude and/or probability of input spike influence on neuronal response (i.e., output spike generation or firing).
  • Efficacy may comprise, for example a parameter 108 (e.g., synaptic weight, delay, probability of transmission, and/or other parameter) by which one or more state variables of the neuron 110 may be changed.
  • Connection efficacy may be changed in accordance with one or more STDP rules.
  • individual connections may utilize connection-specific rules.
  • different classes of connections e.g., fee-forward, lateral, and/or feedback
  • type-specific common STDP rules may be utilized.
  • the STDP rule may comprise an adaptive STDP mechanism that may be determined in real time by a network entity (e.g., the neuron 110 , and/or another entity (connection).
  • a network entity e.g., the neuron 110 , and/or another entity (connection).
  • the adaptive plasticity mechanism may be based on a similarity measure (e.g., a cross-correlogram and/or mutual information) between neuron output (e.g., the output 116 ) and the neuron input (e.g., the input 106 _ 1 , 106 _ 2 , 1063 , 126 in FIG. 1 ).
  • a similarity measure e.g., a cross-correlogram and/or mutual information
  • FIG. 2 illustrates one implementation of such an adaptive mechanism.
  • the cross-correlogram may be determined based on (i) a time record of pre-synaptic input (e.g., the input 106 _ 1 , 106 _ 2 , 106 _ 3 , 126 in FIG. 1 , and/or input 200 , 210 in FIG. 2 ) into the neuron with a time interval t post ⁇ T; and (ii) a time record of post-synaptic output (e.g., the output 116 in FIG.
  • the time interval ⁇ T may be selected form the range between 1 and 100 ms, preferably 40 ins.
  • multiple correlogram estimates (associates with multiple post-synaptic responses 202 , 204 ) may be averaged to produce a time-averaged similarity measure 220 .
  • the average histogram 220 maybe computed by averaging over the last 1000 spikes across all input synapses, or may be computed by averaging over the last 100 second time period.
  • individual correlograms may be constructed for individual connections (e.g., the connections 104 in FIG. 1 ).
  • An averaged (over multiple connections) correlogram may be determined (not shown in FIG. 2 ).
  • individual spikes may be assigned the same amplitude (e.g., binary 1). Accordingly, the binary correlogram 220 may be interpreted as a histogram of pre-synaptic spike occurrence within individual time slots (bins) prior to the post-synaptic response.
  • the time step (bin width) 222 may be selected equal to 1 ms.
  • An averaged similarity measure (e.g., the correlogram 220 ) may be used to construct plasticity rules for the connections of the neuron, as described in detail with respect to the exemplary implementation of FIG. 3 .
  • the STDP rule 300 of FIG. 3 may comprise a long term depression rule 302 and/or a long term potentiation rule 304 .
  • the pre-synaptic portion i.e., where t pre ⁇ t post
  • the post-synaptic portion i.e., where t pre ⁇ t post
  • LTP rule LTP rule
  • the causal portion of the similarity measure (e.g., the portion of the curve 220 of FIG. 2 where t post ⁇ t pre ⁇ 0, also corresponding to the portion of the correlogram where pre-synaptic spikes came before post-synaptic spikes) may be used to construct the LTP rule.
  • the LTP rule may be characterized by a plasticity window Tp 306 in FIG. 3 .
  • values of the similarity measure ( 220 in FIG. 2 ) that fall within the time interval 306 may be used to determine the LTP portion of the plasticity rule.
  • the similarity measure and w( ⁇ t) of plasticity adjustments may be scaled such that the mean value is constrained to be 1, or the maximum value may be constrained to be 1.
  • the LTD portion of the STDP rule may comprise for example a monotonic function of time (e.g., a constant 302 _ 1 , gradually increasing and/or decreasing 302 _ 2 ).
  • the magnitude of the LTD portion may be determined based on a statistical parameter of the LTP portion.
  • the statistical parameter may comprise for instance mean, median, a percentile, maximum, etc.
  • the LTD portion may be characterized by a time window Td 308 .
  • the LTP and LTD windows 306 , 308 may be configured equal to one another.
  • the time window ⁇ T used for determining similarity measure (e.g., the measure 220 in FIG. 2 ) may be set equal to either Tp, Td.
  • the time windows ⁇ T, Tp, Td may also be configured different from one another, such that the window of LTD is longer than LTP (or vice versa), depending on the temporal structure of the pattern to be learned.
  • the adaptively configured STDP rules may be used to adjust efficacy of neuron connections.
  • the adjustment may be performed based on the post-synaptic response of the neuron.
  • the efficacy may comprise synaptic weight.
  • the history of pre-synaptic input may be evaluated.
  • One or more pre-synaptic pulses may be identified within the time window ⁇ T, prior to the post-synaptic spike 230 .
  • Plasticity components e.g., the components 242 , 244 , 246 , corresponding to the pre-synaptic times t pre i of the identified pulses 232 , 234 , 236 , may be combined by an operator 240 to produce weight adjustment w 248 .
  • the adjustment may be performed based on an event, such as timer expiration, buffer overflow, external (reward) indication, and/or other event types).
  • FIGS. 4-6 exemplary implementations of adaptive plasticity methods according to the disclosure are described.
  • the methods of FIGS. 4-6 may be used, for example, for operating the neurons 102 of FIG. 1 .
  • methods of FIG. 4-6 may be implemented in a connection (e.g., the connection 104 , 124 of FIG. 1 ).
  • the methods of FIG. 4-6 may also be implemented in sensory processing apparatus, comprising one or more spiking neuron networks as described with respect to FIG. 8 , infra, thereby advantageously aiding, inter alia improving the speed and or quality of learning, potentially requiring fewer neurons to represent a given input, as well as learning temporally stable patterns.
  • FIG. 4 illustrates a method of determining adaptive plasticity for a spiking neuron based on a similarity measure between neuron input and output, in accordance with one implementation.
  • a neuron generates an output Y.
  • the output may be generated in accordance with a deterministic spike response model, as described for example in co-owned U.S. patent application Ser. No. 13/152,119, entitled “APPARATUS AND METHODS FOR TEMPORALLY PROXIMATE OBJECT RECOGNITION”, filed on Jun. 2, 2011, and patented as U.S. Pat. No. 8,942,466 on Jan. 27, 2015, incorporated herein by reference in its entirety.
  • the spike response process may comprise adaptive threshold adjustment as described in co-owned U.S. patent application Ser. No.
  • the neuron may be operable in accordance with stochastic process, as described in co-owned U.S. patent application Ser. No. 13/487,499, entitled “STOCHASTIC APPARATUS AND METHODS FOR IMPLEMENTING GENERALIZED LEARNING RULES”, filed on Jun. 4, 2012, incorporated herein by reference in its entirety.
  • a history of inputs into the neuron may be accessed.
  • the history may comprise one or more spikes 106 _ 1 , 106 _ 1 , 106 _ 1 in FIG. 1 .
  • the history may comprise for example time data of pre-synaptic spikes stored in a synaptic memory buffer as described in U.S. patent application Ser. No. 13/239,259, entitled “APPARATUS AND METHOD FOR PARTIAL EVALUATION OF SYNAPTIC UPDATES BASED ON SYSTEM EVENTS”, filed on Sep. 21, 2011, and patented as U.S. Pat. No. 8,725,662 on May 13, 2014, incorporated supra.
  • a similarity measure C(X,Y,t) between the output (X) and the input (Y) as a function of time shift (t) may be determined.
  • the similarity measure may use correlation measured at different time shifts; i.e., a correlogram, as described with respect to FIG. 2 , supra.
  • the similarity measure may use for example mutual information at different time shifts as determined as follows:
  • I ⁇ ( X ; Y ) ⁇ y ⁇ Y ⁇ ⁇ x ⁇ X ⁇ p ⁇ ( x , y ) ⁇ log ⁇ ( p ⁇ ( x , y ) p ⁇ ( x ) ⁇ p ⁇ ( y ) ) , where:
  • p(x,y) is the joint probability distribution function of X and Y;
  • p(x) and p(y) are the marginal probability distribution functions of X and Y respectively.
  • a plasticity rule is configured based on the similarity measure.
  • the plasticity rule determination may comprise approach described with respect to FIGS. 2A-3 , supra.
  • FIG. 5 illustrates a method of determining adaptive plasticity for a spiking neuron comprising multiple input connections, in accordance with one implementation.
  • plasticity time window is configured.
  • the LTP and LTD component may comprise the time extent T.
  • the LTP and LTD component may comprise different time extents Tp, Td.
  • the LTD may comprise a constant, a gradually increasing and/or decreasing function (e.g., the functions 302 _ 1 , 302 _ 2 illustrated in FIG. 3 ).
  • an event is detected.
  • the event may comprise post-synaptic spike generation by the neuron.
  • the event may comprise for example an external event (e.g., reinforcement signal), a timer event (e.g., for cyclic updates), a buffer overflow event (e.g., indicative of a memory buffer, storing, for example, pre-synaptic and/or post-synaptic spike history being full or nearly full), etc.
  • a correlogram C i (e.g., between pre-synaptic spike train x i from i th connection and the post-synaptic spike train Y over the time window T) is determined.
  • the correlogram determination comprises:
  • a check may be performed if inputs form other connections need to be processed.
  • the method 500 may return to step 508 .
  • the individual weights may be set to the same value.
  • the weights may be set synapse-specific. For example, the weights could be “0” for very weak synaptic connections, and “1” for strong connections.
  • the LTP portion of the STDP rule (e.g., the rule 304 in FIG. 3 ) is determined using the causal (tpre ⁇ tpost) portion of the average correlogram ⁇ C> of Eqn. 1.
  • the magnitude of the LTD portion of the STDP rule (e.g., the magnitude 310 of the rule 302 _ 1 in FIG. 3 ) is determined.
  • the magnitude may be based on a statistical parameter (e.g., mean, median, percentile, maximum, etc. of the LTP rule portion).
  • FIG. 6 illustrates a method of connection plasticity update based on the adaptive STDP rule, in accordance with one implementation.
  • the update may be based on a post-synaptic spike by the neuron.
  • the update may be based for example on an external event (e.g., reinforcement signal), a timer event (e.g., for cyclic updates), a buffer overflow event (e.g., indicative of a memory buffer, storing, for example, pre-synaptic and/or post-synaptic spike history being full or nearly full), etc.
  • the method proceeds to step 604 , where pre-synaptic update may be performed.
  • the pre-synaptic update may comprise the LTD rule (e.g., the rule determined at step 516 of FIG. 5 and/or rule 302 of FIG. 3 ).
  • the LTP and the LTD rules may or may not be combined to produce the post-rule.
  • the post-rule may be exclusively LTP-based, or may be a combination of both LTP and LTD. The combination may be used in order increase the selectivity of which temporal patterns are potentiated.
  • the post-synaptic update is performed.
  • the post-synaptic update may comprise the combined rule.
  • FIG. 7 illustrates exemplary performance data associated with a learning object recognition apparatus comprising an adaptive plasticity mechanism.
  • a signal processing apparatus e.g., the apparatus 1000 of FIG. 8 described in detail below
  • the input frames may comprise representations of one or more objects, such as the triangles 702 and the squares 704 of frame 700 .
  • the circular panels in FIG. 7 depict receptive fields of the post-synaptic neurons of the network that may develop during learning by the network based on the input (e.g., one or more frames 700 ).
  • the solid dots 718 in FIG. denote responses of the one or more post-synaptic neurons.
  • the receptive fields 710 , 714 , 716 depict performance of a network comprising plasticity mechanism of the prior art. As shown by the receptive fields 710 , 714 , 716 , the network is capable of learning only a single feature (e.g., the triangle 702 ). Furthermore, the prior art plasticity mechanism may cause one or more post-synaptic neurons to generate erroneous responses, as illustrated by the solid circle 718 in FIG. 7 .
  • the receptive fields 720 , 722 , 724 , 726 illustrate that the network (e.g., comprising one or more neurons 110 of FIG. 1 ) is capable of learning both the triangular objects (as shown by the neuron responses in the panels 720 , 722 ) as well as the square objects (as shown by the neuron responses in the panels 724 , 726 ).
  • the performance of the network comprising the exemplary inventive adaptive plasticity mechanism is characterized by a lower number of erroneous responses.
  • FIGS. 7-11 Various exemplary spiking network apparatus implementing one or more of the methods set forth herein (e.g., using the exemplary adaptive mechanisms described above) are now described with respect to FIGS. 7-11 .
  • the illustrated processing apparatus 1000 includes an input interface configured to receive an input sensory signal 1020 .
  • this sensory input comprises electromagnetic waves (e.g., visible light, IR, UV, etc.) entering an imaging sensor array (comprising RGCs, a charge coupled device (CCD), CMOS device, or an active-pixel sensor (APS)).
  • the input signal in this example is a sequence of images (image frames) received from a CCD or CMOS camera via a receiver apparatus, or downloaded from a file.
  • the image may be a two-dimensional matrix of RGB values refreshed at a 24 Hz frame rate.
  • image parameters and components are merely exemplary, and many other image representations (e.g., bitmap, CMYK, grayscale, etc.) and/or frame rates are equally useful with the present disclosure.
  • the apparatus 1000 may also include an encoder 1024 configured to transform (encode) the input signal so as to form an encoded signal 1026 .
  • the encoded signal comprises a plurality of pulses (also referred to as a group of pulses) configured to model neuron behavior.
  • the encoded signal 1026 may be communicated from the encoder 1024 via multiple connections (also referred to as transmission channels, communication channels, or synaptic connections) 1004 to one or more neuronal nodes (also referred to as the detectors) 1002 .
  • FIG. 8 different detectors of the same hierarchical layer are denoted by an “_n” designator, such that e.g., the designator 1002 _ 1 denotes the first detector of the layer 1002 .
  • the encoder can be coupled to any number of detector nodes that is compatible with the detection apparatus hardware and software limitations.
  • a single detector node may be coupled to any practical number of encoders.
  • each of the detectors 1002 _ 1 , 1002 — n contain logic (which may be implemented as a software code, hardware logic, or a combination of thereof) configured to recognize a predetermined pattern of pulses in the encoded signal 1004 , using for example any of the mechanisms described in co-owned U.S. patent application Ser. No. 12/869,573, filed Aug. 26, 2010, entitled “SYSTEMS AND METHODS FOR INVARIANT PULSE LATENCY CODING”, and patent as U.S. Pat. No. 8,315,305 on Nov. 20, 2012, co-owned U.S. patent application Ser. No. 12/869,583, filed Aug.
  • the designators 1008 _ 1 , 1008 — n denote output of the detectors 1002 _ 1 , 1002 — n , respectively.
  • the detection signals are delivered to a next layer of the detectors 1012 (comprising detectors 1012 — l , 1012 — m , 1012 — k ) for recognition of complex object features and objects, similar to the exemplary configuration described in commonly owned and co-pending U.S. patent application Ser. No. 13/152,084, filed Jun. 2, 2011, entitled “APPARATUS AND METHODS FOR PULSE-CODE INVARIANT OBJECT RECOGNITION”, incorporated herein by reference in its entirety.
  • each subsequent layer of detectors is configured to receive signals from the previous detector layer, and to detect more complex features and objects (as compared to the features detected by the preceding detector layer). For example, a bank of edge detectors is followed by a bank of bar detectors, followed by a bank of corner detectors, and so on, thereby enabling alphabet recognition by the apparatus.
  • Each of the detectors 1002 may output detection (post-synaptic) signals on communication channels 1008 _ 1 , 1008 — n (with appropriate latency) that may propagate with different conduction delays to the detectors 1012 .
  • the detector cascade of the apparatus of FIG. 8 may contain any practical number of detector nodes and detector banks determined, inter alia, by the software/hardware resources of the detection apparatus and complexity of the objects being detected.
  • the sensory processing apparatus implementation illustrated in FIG. 8 may further comprise lateral connections 1006 .
  • the connections 1006 are configured to communicate post-synaptic activity indications between neighboring neurons of the same hierarchy level, as illustrated by the connection 1006 _ 1 in FIG. 8 .
  • the neighboring neuron may comprise neurons having overlapping inputs (e.g., the inputs 1004 _ 1 , 1004 — n in FIG. 8 ), so that the neurons may compete in order to not learn the same input features.
  • the neighboring neurons may comprise spatially proximate neurons such as being disposed within a certain volume/area from one another on a 3-dimensional (3D) and or two-dimensional (2D) space.
  • the apparatus 1000 may also comprise feedback connections 1014 , configured to communicate context information from detectors within one hierarchy layer to previous layers, as illustrated by the feedback connections 1014 _ 1 in FIG. 8 .
  • the feedback connection 1014 _ 2 is configured to provide feedback to the encoder 1024 thereby facilitating sensory input encoding, as described in detail in commonly owned and co-pending U.S. patent application Ser. No. 13/152,084, filed Jun. 2, 2011, entitled “APPARATUS AND METHODS FOR PULSE-CODE INVARIANT OBJECT RECOGNITION”, incorporated supra.
  • FIG. 9 One particular implementation of the computerized neuromorphic processing system, adapted for operating a computerized spiking network (and implementing the exemplary conditional plasticity methodology described supra), is illustrated in FIG. 9 .
  • the computerized system 1100 of FIG. 9 comprises an input interface 1110 , such as for example an image sensor, a computerized spiking retina, an audio array, a touch-sensitive input device, etc.
  • the input interface 1110 is coupled to the processing block (e.g., a single or multi-processor block) via the input communication interface 1114 .
  • the system 1100 further comprises a random access memory (RAM) 1108 , configured to store neuronal states and connection parameters (e.g., weights 108 in FIG. 1 ), and to facilitate synaptic updates.
  • RAM random access memory
  • synaptic updates are performed according to the description provided in, for example, in U.S. patent application Ser. No. 13/239,255 filed Sep. 21, 2011, entitled “APPARATUS AND METHODS FOR SYNAPTIC UPDATE IN A PULSE-CODED NETWORK”, incorporated by reference supra.
  • the memory 1108 is coupled to the processor 1102 via a direct connection (memory bus) 1116 .
  • the memory 1108 may also be coupled to the processor 1102 via a high-speed processor bus 1112 ).
  • the system 1100 may further comprise a nonvolatile storage device 1106 , comprising, inter alia, computer readable instructions configured to implement various aspects of spiking neuronal network operation (e.g., sensory input encoding, connection plasticity, operation model of neurons, etc.).
  • the nonvolatile storage 1106 may be used for instance to store state information of the neurons and connections when, for example, saving/loading network state snapshot, or implementing context switching (e.g., saving current network configuration (comprising, inter alia, connection weights and update rules, neuronal states and learning rules, etc.) for later use, and loading of a previously stored network configuration.
  • the computerized apparatus 1100 is coupled to one or more external processing/storage/input devices via an I/O interface 1120 , such as a computer I/O bus (PCI-E), wired (e.g., Ethernet) or wireless (e.g., Wi-Fi) network connection.
  • I/O interface 1120 such as a computer I/O bus (PCI-E), wired (e.g., Ethernet) or wireless (e.g., Wi-Fi) network connection.
  • PCI-E computer I/O bus
  • wired e.g., Ethernet
  • wireless e.g., Wi-Fi
  • the input/output interface comprises a speech input (e.g., a microphone) and a speech recognition module configured to receive and recognize user commands.
  • a speech input e.g., a microphone
  • a speech recognition module configured to receive and recognize user commands.
  • processing devices may be used with computerized system 1100 , including but not limited to, a single core/multicore CPU, DSP, FPGA, GPU, ASIC, combinations thereof, and/or other processors.
  • Various user input/output interfaces are similarly applicable to implementations of the disclosure including, for example, an LCD/LED monitor, touch-screen input and display device, speech input device, stylus, light pen, trackball, end the likes.
  • the neuromorphic processing system 1130 of FIG. 10 comprises a plurality of processing blocks (micro-blocks) 1140 , where each micro-block comprises a computing logic core 1132 and a memory block 1134 .
  • the logic core 1132 is configured to implement various aspects of neuronal node operation, such as the node model, and synaptic update rules (e.g., the I-STDP) and/or other tasks relevant to network operation.
  • the memory block is configured to store, inter alia, neuronal state variables and connection parameters (e.g., weights, delays, I/O mapping) of connections 1138 .
  • connection layout in FIG. 10 is exemplary, and many other connection implementations (e.g., one to all, all to all, etc.) are compatible with the disclosure.
  • the neuromorphic apparatus 1130 is configured to receive input (e.g., visual input) via the interface 1142 .
  • input e.g., visual input
  • the apparatus 1130 may provide feedback information via the interface 1142 to facilitate encoding of the input signal.
  • the neuromorphic apparatus 1130 is configured to provide output (e.g., an indication of recognized object or a feature, or a motor command, e.g., to zoom/pan the image array) via the interface 1144 .
  • output e.g., an indication of recognized object or a feature, or a motor command, e.g., to zoom/pan the image array
  • the apparatus 1130 may interface to external fast response memory (e.g., RAM) via high bandwidth memory interface 1148 , thereby enabling storage of intermediate network operational parameters (e.g., spike timing, etc.).
  • external fast response memory e.g., RAM
  • high bandwidth memory interface 1148 thereby enabling storage of intermediate network operational parameters (e.g., spike timing, etc.).
  • the apparatus 1130 may also interface to external slower memory (e.g., Flash, or magnetic (hard drive)) via lower bandwidth memory interface 1146 , in order to facilitate program loading, operational mode changes, and retargeting, where network node and connection information for a current task is saved for future use and flushed, and previously stored network configuration is loaded in its place.
  • FIG. 11 A illustrates implementations of a shared bus neuromorphic computerized system comprising micro-blocks 1140 , described with respect to FIG. 10 , supra, coupled to a shared interconnect.
  • the apparatus 1145 of FIG. 11A utilizes one (or more) shared bus(es) 1146 in order to interconnect micro-blocks 1140 with one another.
  • FIG. 11B illustrates one implementation of cell-based neuromorphic computerized system architecture configured to implement Conditional plasticity mechanism in a spiking network.
  • the neuromorphic system 1150 of FIG. 11B comprises a hierarchy of processing blocks (cells block).
  • the lowest level L1 cell 1152 of the apparatus 1150 may comprise logic and memory, and may be configured similar to the micro block 1140 of the apparatus shown in FIG. 10 .
  • a number of cell blocks may be arranges in a cluster and communicate with one another a local interconnects 1162 , 1164 . Each such cluster may form a higher-level cell, e.g., cell L2, denoted as 1154 in FIG. 11B .
  • L2 clusters may communicate with one another via a second-level interconnect 1166 and form a super-cluster L3, denoted as 1156 in FIG. 11B .
  • the super-clusters 1154 may for example communicate via a third level interconnect 1168 , and may form a next level cluster, and so on.
  • the hierarchical structure of the apparatus 1150 comprising a given number (e.g., four) cells per level, is merely one exemplary implementation, and other implementations may comprise more or fewer cells per level, and/or fewer or more levels, as well as yet other types of architectures.
  • Different cell levels e.g., L1, L2, L3 of the exemplary apparatus 1150 of FIG. 11B may be configured to perform functionality with various levels of complexity.
  • different L1 cells may process in parallel different portions of the visual input (e.g., encode different frame macro-blocks), with the L2, L3 cells performing progressively higher-level functionality (e.g., edge detection, object detection).
  • Different L2, L3 cells may also perform different aspects of operating for example a robot, with one or more L2/L3 cells processing visual data from a camera, and other L2/L3 cells operating a motor control block for implementing lens motion when e.g., tracking an object, or performing lens stabilization functions.
  • the neuromorphic apparatus 1150 may receive input (e.g., visual input) via the interface 1160 .
  • input e.g., visual input
  • the apparatus 1150 may provide feedback information via the interface 1160 to facilitate encoding of the input signal.
  • the neuromorphic apparatus 1150 may provide output (e.g., an indication of recognized object or a feature, or a motor command, e.g., to zoom/pan the image array) via the interface 1170 .
  • the apparatus 1150 may perform all of the I/O functionality using single I/O block (not shown).
  • the apparatus 1150 may also interface to external fast response memory (e.g., RAM) via high bandwidth memory interface (not shown), thereby enabling storage of intermediate network operational parameters (e.g., spike timing, etc.).
  • external fast response memory e.g., RAM
  • high bandwidth memory interface not shown
  • intermediate network operational parameters e.g., spike timing, etc.
  • external slower memory e.g., flash, or magnetic (hard drive)
  • lower bandwidth memory interface not shown
  • the plasticity mechanism described herein may advantageously improve the quality of the feature learning (as characterized by a number of false positives) and/or to increase learning speed.
  • the adaptive plasticity mechanism may enable the processing apparatus to learn temporally stable patterns.
  • Various aspects of the disclosure may advantageously be applied to, inter alia, the design and operation of large spiking neural networks configured to process streams of input stimuli, in order to aid in detection and functional binding related aspect of the input.
  • conditional plasticity mechanisms described herein may be implemented in a spiking neuron of a network, or in a connection of the network.
  • the approach of the disclosure can advantageously, among other things, achieve faster learning and improve the quality of learned features and (iv) enable efficient network operation (e.g., visual input encoding) for a wide variety of input conditions. It will be appreciated that the increased network stability and flexibility may be traded for (a) a less complex, less costly and more robust network capable of processing the same feature set with fewer neurons; and/or (b) a more capable, higher performance network capable of processing larger and more complex feature set with the same number of neurons, when compared to the prior art solutions.
  • exemplary implementations of the various aspects of the present innovation are useful in a variety of devices including without limitation prosthetic devices, autonomous and robotic apparatus, and other electromechanical devices requiring sensory processing functionality.
  • robotic devises are manufacturing robots (e.g., automotive), military, medical (e.g. processing of microscopy, x-ray, ultrasonography, tomography).
  • autonomous vehicles include rovers, unmanned air vehicles, underwater vehicles, smart appliances (e.g. ROOMBA®), etc.
  • Implementations of the principles of the disclosure are applicable to video data compression and processing in a wide variety of stationary and portable devices, such as, for example, smart phones, portable communication devices, notebook, netbook and tablet computers, surveillance camera systems, and practically any other computerized device configured to process vision data
  • Implementations of the principles of the disclosure are further applicable to a wide assortment of applications including computer human interaction (e.g., recognition of gestures, voice, posture, face, etc.), controlling processes (e.g., an industrial robot, autonomous and other vehicles), augmented reality applications, organization of information (e.g., for indexing databases of images and image sequences), access control (e.g., opening a door based on a gesture, opening an access way based on detection of an authorized person), detecting events (e.g., for visual surveillance or people or animal counting, tracking), data input, financial transactions (payment processing based on recognition of a person or a special payment symbol) and many others.
  • computer human interaction e.g., recognition of gestures, voice, posture, face, etc.
  • controlling processes e.g., an industrial robot, autonomous and other vehicles
  • augmented reality applications organization of information (e.g., for indexing databases of images and image sequences)
  • access control e.g., opening a door based on a gesture, opening an
  • the disclosure can be used to simplify tasks related to motion estimation, such as where an image sequence is processed to produce an estimate of the object position (and hence velocity) either at each points in the image or in the 3D scene, or even of the camera that produces the images.
  • tasks related to motion estimation such as where an image sequence is processed to produce an estimate of the object position (and hence velocity) either at each points in the image or in the 3D scene, or even of the camera that produces the images.
  • ego motion i.e., determining the three-dimensional rigid motion (rotation and translation) of the camera from an image sequence produced by the camera; following the movements of a set of interest points or objects (e.g., vehicles or humans) in the image sequence and with respect to the image plane.
  • portions of the object recognition system are embodied in a remote server, comprising a computer readable apparatus storing computer executable instructions configured to perform pattern recognition in data streams for various applications, such as scientific, geophysical exploration, surveillance, navigation, data mining (e.g., content-based image retrieval).
  • applications such as scientific, geophysical exploration, surveillance, navigation, data mining (e.g., content-based image retrieval).
  • applications such as scientific, geophysical exploration, surveillance, navigation, data mining (e.g., content-based image retrieval).

Abstract

Apparatus and methods for plasticity in a spiking neuron network. In one implementation, a plasticity mechanism is configured based on a similarity measure between neuron post-synaptic and pre-synaptic activity. The similarity measure may comprise a cross-correlogram between the output spike train and input spike train, determined over a plasticity window. Several correlograms, corresponding to individual input connections delivering pre-synaptic input, may be combined. The combination may comprise for example a weighted average. The averaged correlograms may be used to construct the long term potentiation component of the plasticity. The long term depression component of the plasticity may comprise e.g., a monotonic function based on a statistical parameter associated with the adaptively determined long term potentiation component.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS
This application is related to co-owned U.S. patent application Ser. No. 13/152,119, entitled “SENSORY INPUT PROCESSING APPARATUS AND METHODS”, filed on Jun. 2, 2011 and patented as U.S. Pat. No. 8,942,466 on Jan. 27, 2015, co-owned and co-pending U.S. patent application Ser. No. 13/465,924, entitled “SPIKING NEURAL NETWORK FEEDBACK APPARATUS AND METHODS”, filed May 7, 2012, co-owned and co-pending U.S. patent application Ser. No. 13/465,903 entitled “SENSORY INPUT PROCESSING APPARATUS IN A SPIKING NEURAL NETWORK”, filed May 7, 2012, a co-owned and co-pending U.S. patent application Ser. No. 13/465,918, entitled “SPIKING NEURAL NETWORK OBJECT RECOGNITION APPARATUS AND METHODS”, filed May 7, 2012, co-owned U.S. patent application Ser. No. 13/488,106, entitled “SPIKING NEURON NETWORK APPARATUS AND METHODS”, filed Jun. 4, 2012, co-owned and co-pending U.S. patent application Ser. No. 13/488,144, entitled “SPIKING NEURON NETWORK APPARATUS AND METHODS”, filed Jun. 4, 2012, co-owned, co-pending U.S. patent application Ser. No. 13/541,531, entitled “CONDITIONAL PLASTICITY SPIKING NEURON NETWORK APPARATUS AND METHODS”, filed Jul. 3, 2012, co-owned, co-pending U.S. patent application Ser. No. 13/660,945, entitled “MODULATED PLASTICITY APPARATUS AND METHODS FOR SPIKING NEURON NETWORKS”, filed on Oct. 25, 2012, co-owned U.S. patent application Ser. No. 13/660,967, entitled “APPARATUS AND METHODS FOR ACTIVITY-BASED PLASTICITY IN A SPIKING NEURON NETWORK”, filed on Oct. 25, 2012 and patented as U.S patent No. 8,972,315 on Mar. 3, 2015, and co-owned, co-pending U.S. patent application Ser. No. 13/660,982, entitled “SPIKING NEURON SENSORY PROCESSING APPARATUS AND METHODS FOR SALIENCY DETECTION”, filed on Oct. 25, 2012, each of the foregoing incorporated herein by reference in its entirety.
COPYRIGHT
A portion of the disclosure of this patent document contains material that is subject to copyright protection. The copyright owner has no objection to the facsimile reproduction by anyone of the patent document or the patent disclosure, as it appears in the Patent and Trademark Office patent files or records, but otherwise reserves all copyright rights whatsoever.
BACKGROUND
1. Technological Field
The present disclosure relates generally to artificial neural networks, and more particularly in one exemplary aspect to computer apparatus and methods for plasticity implementation in a pulse-code neural network.
2. Description of Related Art
Artificial spiking neural networks are frequently used to gain an understanding of biological neural networks, and for solving artificial intelligence problems. These networks typically employ a pulse-coded mechanism, which encodes information using timing of the pulses. Such pulses (also referred to as “spikes” or ‘impulses’) are short-lasting (typically on the order of 1-2 ms) discrete temporal events. Several exemplary embodiments of such encoding are described in commonly owned and co-pending U.S. patent application Ser. No. 13/152,084 entitled “APPARATUS AND METHODS FOR PULSE-CODE INVARIANT OBJECT RECOGNITION”, filed Jun. 2, 2011, and co-owned U.S. patent application Ser. No. 13/152,119,filed Jun. 2, 2011, entitled “SENSORY INPUT PROCESSING APPARATUS AND METHODS”, and patented as U.S. Pat. No. 8,942,466 on Jan. 27, 2015, each incorporated herein by reference in its entirety.
Typically, artificial spiking neural networks, such as the exemplary network described in owned U.S. patent application Ser. No. 13/541,531, entitled “CONDITIONAL PLASTICITY SPIKING NEURON NETWORK APPARATUS AND METHODS”, may comprise a plurality of units (or nodes), which correspond to neurons in a biological neural network. Any given unit may be connected to many other units via connections, also referred to as communications channels, and/or synaptic connections. The units providing inputs to any given unit are commonly referred to as the pre-synaptic units, while the unit receiving the inputs is referred to as the post-synaptic unit.
Each of the unit-to-unit connections may be assigned, inter alia, a connection efficacy, which in general may refer to a magnitude and/or probability of input spike influence on unit output response (i.e., output spike generation/firing). The efficacy may comprise, for example a parameter—synaptic weight—by which one or more state variables of post-synaptic unit are changed. During operation of a pulse-code network, synaptic weights may be dynamically adjusted using what is referred to as the spike-timing dependent plasticity (STDP) in order to implement, among other things, network learning. In some implementations, larger weights may be associated with a greater effect a synapse has on the activity of the post-synaptic neuron.
In some existing plasticity implementations, connections that deliver inputs (to a given unit) prior to generation of post-synaptic response may be potentiated, while connections that deliver inputs after the generation of the post-synaptic response may be depressed. The choice of plasticity functional dependence may determine network behavior. Accordingly, various implementations plasticity mechanisms exist including, for example, the use of target connection efficiency (that may be defined as a ratio of a number of input (pre-synaptic) spikes Nfire delivered to a neuron via the connection that are followed by neuron response (e.g., post-synaptic spike) generation, to the total number of input spikes Ntot delivered to the neuron via the connection. However, existing plasticity implementations do not always provide for network behavior, particularly when input characteristics change.
Consequently there is a salient need for improved adaptive plasticity mechanisms to enable a spiking neuron network capable of operating in a wide variety of input and network dynamic regimes.
SUMMARY OF THE DISCLOSURE
The present disclosure satisfies the foregoing needs by providing, inter alia, apparatus and methods for implementing adaptive plasticity in spiking neuron networks that may be dynamically adjusted in accordance with the connection activity thereby enhancing the learning performance of a neural network.
In a first aspect of the disclosure, a method of updating an efficacy of a connection configured to communicate an input to an artificial spiking neuron is disclosed. In one embodiment, the method includes: determining a similarity measure between the input and a neuron output; identifying one or more components associated with one or more lag values, respectively, from the similarity measure; and updating the efficacy based on the one or more components.
In a second aspect, computerized spiking neuron apparatus is disclosed. In one embodiment, the apparatus comprise a storage medium, the storage medium comprising a plurality of executable instructions configured to, when executed, adjust an efficacy of an interface of the neuron by at least: adjustment of neuron excitability based on one or more pre-synaptic inputs; storage of one or more pre-synaptic times associated with one or more pre-synaptic inputs; and when the excitability is above a threshold: generation of post-synaptic output; determination of a similarity measure between the neuron output signal, comprising the response, and the one or more pre-synaptic inputs; determination of a long term potentiation component (LTP) of the adaptive plasticity based on the similarity measure; and determination of a long term depression component of the adaptive plasticity based on a parameter associated with the LTP
In one variant, the response is generated by the neuron based on the one or more pre-synaptic inputs.
In another aspect, a computerized spiking neuron network system is disclosed. In one embodiment, the system is configured to implement an adaptive plasticity in the network, and includes: one or more processors configured to execute computer program modules, the execution of the computer program modules configured to cause the one or more processors, based on a response by a neuron of the network, to adjust one or more connections capable of providing input to the neuron by at least: determination of one or more similarity measures between the neuron output signal, comprising the response, and one or more input signals provided by the one or more connections, respectively; determination of a long term potentiation component (LTP) of the adaptive plasticity based on an average of the one or more similarity measures; and determination of a long term depression component of the adaptive plasticity based on a parameter associated with the LTP.
In another aspect, computer readable apparatus is disclosed. In one embodiment, the apparatus includes at least one computer program configured to, when executed, implement an adaptive plasticity in a neural network.
Further features of the present disclosure, its nature and various advantages will be more apparent from the accompanying drawings and the following detailed description.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a block diagram depicting an artificial spiking neural network according to one implementation.
FIG. 2 is a graphical illustration depicting histogram determination based on post-synaptic and pre-synaptic activity of a unit of the spiking network of FIG. 1, according to one implementation.
FIG. 3 is a plot depicting adaptively constructed spike-time dependent plasticity (STDP) for use in the spiking network of FIG. 1, according to one implementation.
FIG. 4 is a logical flow diagram illustrating determination of an adaptive plasticity mechanism, in accordance with one implementation.
FIG. 5 is a logical flow diagram illustrating a method of determining adaptive plasticity for a spiking neuron comprising multiple input connections, in accordance with one implementation.
FIG. 6 is a logical flow diagram illustrating a method of connection plasticity update based on the adaptive STDP rule, in accordance with one implementation.
FIG. 7 is a block diagram illustrating performance of a sensory processing apparatus comprising adaptive plasticity mechanism in accordance with one implementation.
FIG. 8 is a block diagram illustrating a sensory processing apparatus comprising adaptive plasticity mechanism in accordance with one implementation.
FIG. 9 is a block diagram illustrating a computerized system useful for, inter alia, providing an adaptive plasticity mechanism in a spiking network, in accordance with one implementation.
FIG. 10 is a block diagram illustrating a neuromorphic computerized system useful with, inter alia, adaptive plasticity mechanism in a spiking network, in accordance with one implementation.
FIG. 11A is a block diagram illustrating a hierarchical neuromorphic computerized system architecture useful with, inter alia, adaptive plasticity mechanism in a spiking network, in accordance with one implementation.
FIG. 11B is a block diagram illustrating a cell-type neuromorphic computerized system architecture useful with, inter alia, adaptive plasticity mechanism in a spiking network, in accordance with one implementation.
All Figures disclosed herein are © Copyright 2012 Brain Corporation. All rights reserved.
DETAILED DESCRIPTION
Embodiments and implementations of the various aspects of the present innovation will now be described in detail with reference to the drawings, which are provided as illustrative examples so as to enable those skilled in the art to practice the disclosure. Notably, the figures and examples below are not meant to limit the scope of the present disclosure to a single embodiment or implementation, but other embodiments and implementations are possible by way of interchange of or combination with some or all of the described or illustrated elements. Wherever convenient, the same reference numbers will be used throughout the drawings to refer to same or like parts.
Where certain elements of these embodiments or implementations can be partially or fully implemented using known components, only those portions of such known components that are necessary for an understanding of the present disclosure will be described, and detailed descriptions of other portions of such known components will be omitted so as not to obscure the innovation.
In the present specification, an embodiment or implementations showing a singular component should not be considered limiting; rather, the invention is intended to encompass other embodiments or implementations including a plurality of the same component, and vice-versa, unless explicitly stated otherwise herein.
Further, the present invention encompasses present and future known equivalents to the components referred to herein by way of illustration.
As used herein, the term “bus” is meant generally to denote all types of interconnection or communication architecture that is used to access the synaptic and neuron memory. The “bus” could be optical, wireless, infrared or another type of communication medium. The exact topology of the bus could be for example standard “bus”, hierarchical bus, network-on-chip, address-event-representation (AER) connection, or other type of communication topology used for accessing, e.g., different memories in pulse-based system.
As used herein, the terms “computer”, “computing device”, and “computerized device”, include, but are not limited to, personal computers (PCs) and minicomputers, whether desktop, laptop, or otherwise, mainframe computers, workstations, servers, personal digital assistants (PDAs), handheld computers, embedded computers, programmable logic device, personal communicators, tablet computers, portable navigation aids, J2ME equipped devices, cellular telephones, smart phones, personal integrated communication or entertainment devices, or literally any other device capable of executing a set of instructions and processing an incoming data signal.
As used herein, the term “computer program” or “software” is meant to include any sequence or human or machine cognizable steps which perform a function. Such program may be rendered in virtually any programming language or environment including, for example, C/C++, C#, Fortran, COBOL, MATLAB™, PASCAL, Python, assembly language, markup languages (e.g., HTML, SGML, XML, VoXML), and the like, as well as object-oriented environments such as the Common Object Request Broker Architecture (CORBA), Java™ (including J2ME, Java Beans, etc.), Binary Runtime Environment (e.g., BREW), and the like.
As used herein, the terms “connection”, “link”, “synaptic channel”, “transmission channel”, “delay line”, are meant generally to denote a causal link between any two or more entities (whether physical or logical/virtual), which enables information exchange between the entities.
As used herein, the term “memory” includes any type of integrated circuit or other storage device adapted for storing digital data including, without limitation, ROM. PROM, EEPROM, DRAM, Mobile DRAM, SDRAM, DDR/2 SDRAM, EDO/FPMS, RLDRAM, SRAM, “flash” memory (e.g., NAND/NOR), memristor memory, and PSRAM.
As used herein, the terms “processor”, “microprocessor” and “digital processor” are meant generally to include all types of digital processing devices including, without limitation, digital signal processors (DSPs), reduced instruction set computers (RISC), general-purpose (CISC) processors, microprocessors, gate arrays (e.g., field programmable gate arrays (FPGAs)), PLDs, reconfigurable computer fabrics (RCFs), array processors, secure microprocessors, and application-specific integrated circuits (ASICs). Such digital processors may be contained on a single unitary IC die, or distributed across multiple components.
As used herein, the term “network interface” refers to any signal, data, or software interface with a component, network or process including, without limitation, those of the FireWire (e.g., FW400, FW800, etc.), USB (e.g., USB2), Ethernet (e.g., 10/100, 10/100/1000 (Gigabit Ethernet), 10-Gig-E, etc.), MoCA, Coaxsys (e.g., TVnet™), radio frequency tuner (e.g., in-band or OOB, cable modem, etc.), Wi-Fi (802.11), WiMAX (802.16), PAN (e.g., 802.15), cellular (e.g., 3G, LTE/LTE-A/TD-LTE, GSM, etc.) or IrDA families.
As used herein, the terms “pulse”, “spike”, “burst of spikes”, and “pulse train” are meant generally to refer to, without limitation, any type of a pulsed signal, e.g., a rapid change in some characteristic of a signal, e.g., amplitude, intensity, phase or frequency, from a baseline value to a higher or lower value, followed by a rapid return to the baseline value and may refer to any of a single spike, a burst of spikes, an electronic pulse, a pulse in voltage, a pulse in electrical current, a software representation of a pulse and/or burst of pulses, a software message representing a discrete pulsed event, and any other pulse or pulse type associated with a discrete information transmission system or mechanism.
As used herein, the term “receptive field” is used to describe sets of weighted inputs from filtered input elements, where the weights may be adjusted.
As used herein, the term “Wi-Fi” refers to, without limitation, any of the variants of IEEE-Std. 802.11 or related standards including 802.11 a/b/g/n/s/v and 802.11-2012.
As used herein, the term “wireless” means any wireless signal, data, communication, or other interface including without limitation Wi-Fi, Bluetooth, 3G (3GPP/3GPP2), HSDPA/HSUPA, TDMA, CDMA (e.g., IS-95A, WCDMA, etc.), FHSS, DSSS, GSM, PAN/802.15, WiMAX (802.16), 802.20, narrowband/FDMA, OFDM, PCS/DCS, LTE/LTE-A/TD-LTE, analog cellular, CDPD, RFID or NFC (e.g., EPC Global Gen. 2, ISO 14443, ISO 18000-3), satellite systems, millimeter wave or microwave systems, acoustic, and infrared (e.g., IrDA).
Overview
The present disclosure provides, in one salient aspect, apparatus and methods for implementing adaptive plasticity mechanisms configured to, inter alia, to improve learning (faster learning or learn more useful features) in artificial spiking neuron networks, without requiring pre-defined and/or hard coded plasticity rules.
In one or more implementations, the adaptive plasticity mechanism may be based on a similarity measure (e.g., a cross-correlogram, cross-correlation, convolution, deconvolution, and/or mutual information) between neuron output and neuron input. When the neuron generates an output (fires a spike), the cross-correlogram may be determined based on (i) a time record of pre-synaptic input into the neuron with a time interval ΔT prior to the output; and (ii) a time record of post-synaptic output by the neuron with the same time interval. In some implementations of neurons comprising multiple pre-synaptic connections (physical and/or logical), individual correlograms may be constructed for individual connections. An averaged (over multiple connections) correlogram may be determined. In some implementations, multiple correlogram estimates (associates with multiple post-synaptic responses) may be averaged to produce a time-averaged similarity measure. The averaging may be implemented in only time with individual synapses having their respective correlogram. In some implementations, the averaging may be implemented over both time and synapses so that a single correlogram may be shared by all input synapses.
In some implementations, individual spikes may be assigned the same amplitude (e.g., binary 1). Accordingly, the binary correlogram may be interpreted as a histogram of pre-synaptic spike occurrence within individual time slots (bins) prior to the post-synaptic response.
Averaged similarity measure may be used to construct plasticity rules for the connections of the neuron. The STDP rule may comprise for instance a long term depression (LTD) rule and/or a long term potentiation (LTP) rule.
The causal portion (i.e., the portion of the similarity measure for which the pre-synaptic spikes occur before the post spikes) of the similarity measure may be used, inter alia, to construct the LTP rule. Magnitude of plasticity adjustments may be scaled such that the mean of the similarity measure is a constant. In some implementations, magnitude of plasticity adjustments may be scaled such that the maximum value of the similarity measure is a constant.
In some implementations, The LTD portion of the STDP rule may comprise a monotonic function (e.g., a constant, gradually increasing and/or decreasing). Magnitude of the LTD portion may be determined based on a statistical parameter of the LTP portion. In some implementations, the statistical parameter may comprise maximum, mean, and/or median of the correlogram).
The adaptively configured STDP rules may be used to adjust weights of neuron input connections. In some implementations, the adjustment may be performed based on the post-synaptic response by the neuron.
The plasticity methodology describe herein may advantageously improve learning of the neural network. The improved learning may be characterized by a faster convergence, convergence to more meaningful features, fewer recognition errors, requiring fewer neurons to represent a given input. In some implementations, the activity bases plasticity mechanism may enable learning of temporally stable patterns (e.g., learning a component of the temporal feature without drift).
In another aspect of the disclosure, adaptive adjustment methodologies are used to implement processing of visual sensory information and feature/object recognition using spiking neuronal networks. Portions of the object recognition apparatus can be embodied for example in a remote computerized apparatus (e.g., server), comprising a computer readable apparatus.
Embodiments of the foregoing plasticity mechanism of the present disclosure are useful in a variety of applications including for instance a prosthetic device, autonomous robotic apparatus, and other electromechanical devices requiring visual or other sensory data processing functionality.
Methods
Detailed descriptions of the various embodiments and implementations of the apparatus and methods of the disclosure are now provided. Although certain aspects of the disclosure can best be understood in the context of the visual and sensory information processing using spiking neural networks, the disclosure is not so limited, and implementations of the disclosure may also be used in a wide variety of other applications, including for instance in implementing connection adaptation in pulse-code neural networks.
Implementations of the disclosure may be for example deployed in a hardware and/or software realization of a neuromorphic computer system. In one such implementation, a robotic system may include a processor embodied in an application specific integrated circuit, which can be adapted or configured for use in an embedded application (such as a prosthetic device).
FIG. 1 illustrates one exemplary implementation of a spiking neuronal network of the disclosure, configured to process sensory information using adaptive plasticity mechanism. The network 100 may comprise one or more spiking neurons, e.g., the neuron 110 in FIG. 1), The neuron 110 may be configured to receive feed-forward spiking input via connections 104. In some implementations, the neuron 110 may be configured to receive feedback spiking input via the connections 124. The neuron 110 may generate output (e.g., a post-synaptic spike) using any of applicable methodologies such as for example those described in co-owned and co-pending U.S. patent application Ser. No. 13/152,105 filed on Jun. 2, 2011, and entitled “APPARATUS AND METHODS FOR TEMPORALLY PROXIMATE OBJECT RECOGNITION”, incorporated by reference herein in its entirety. The output spikes of the neuron 110 may be propagated via the connection 114. Post-synaptic spike generation is well-established in the spiking network arts, and accordingly will not be described in detail herein for brevity and clarity of presentation of the inventive aspects of the present disclosure.
The connections 104_1, 104_2, 104_3, 124 may be characterized by connection efficacy. Efficacy may refer to a magnitude and/or probability of input spike influence on neuronal response (i.e., output spike generation or firing). Efficacy may comprise, for example a parameter 108 (e.g., synaptic weight, delay, probability of transmission, and/or other parameter) by which one or more state variables of the neuron 110 may be changed.
Connection efficacy may be changed in accordance with one or more STDP rules. In some implementations, individual connections may utilize connection-specific rules. In one or more implementations, different classes of connections (e.g., fee-forward, lateral, and/or feedback) may utilize type-specific common STDP rules.
In some implementations, the STDP rule may comprise an adaptive STDP mechanism that may be determined in real time by a network entity (e.g., the neuron 110, and/or another entity (connection).
In one or more implementations, the adaptive plasticity mechanism may be based on a similarity measure (e.g., a cross-correlogram and/or mutual information) between neuron output (e.g., the output 116) and the neuron input (e.g., the input 106_1, 106_2, 1063, 126 in FIG. 1).
FIG. 2 illustrates one implementation of such an adaptive mechanism. When the neuron generates an output (fires a spike 116_1, and/or spike 202 in FIG. 2) at time tpost, the cross-correlogram may be determined based on (i) a time record of pre-synaptic input (e.g., the input 106_1, 106_2, 106_3, 126 in FIG. 1, and/or input 200, 210 in FIG. 2) into the neuron with a time interval tpost−ΔT; and (ii) a time record of post-synaptic output (e.g., the output 116 in FIG. 1) by the neuron with the same time interval. The time interval ΔT may be selected form the range between 1 and 100 ms, preferably 40 ins. In some implementations, multiple correlogram estimates (associates with multiple post-synaptic responses 202, 204) may be averaged to produce a time-averaged similarity measure 220. In some implementations, the average histogram 220 maybe computed by averaging over the last 1000 spikes across all input synapses, or may be computed by averaging over the last 100 second time period.
In some implementations where the neuron receives multiple pre-synaptic connections (physical and/or logical, as illustrated in FIG. 1), individual correlograms may be constructed for individual connections (e.g., the connections 104 in FIG. 1). An averaged (over multiple connections) correlogram may be determined (not shown in FIG. 2).
In some implementations, individual spikes (e.g., the spike groups 200, 210 in FIG. 2) may be assigned the same amplitude (e.g., binary 1). Accordingly, the binary correlogram 220 may be interpreted as a histogram of pre-synaptic spike occurrence within individual time slots (bins) prior to the post-synaptic response. In some implementations, the time step (bin width) 222 may be selected equal to 1 ms.
An averaged similarity measure (e.g., the correlogram 220) may be used to construct plasticity rules for the connections of the neuron, as described in detail with respect to the exemplary implementation of FIG. 3.
The STDP rule 300 of FIG. 3 may comprise a long term depression rule 302 and/or a long term potentiation rule 304. For example, the pre-synaptic portion (i.e., where tpre<tpost) may comprise the LTD rule, and the post-synaptic portion (i.e., where tpre≧tpost) may comprise the LTP rule, as shown in FIG. 3.
The causal portion of the similarity measure (e.g., the portion of the curve 220 of FIG. 2 where tpost−tpre<0, also corresponding to the portion of the correlogram where pre-synaptic spikes came before post-synaptic spikes) may be used to construct the LTP rule. In some implementations, the LTP rule may be characterized by a plasticity window Tp 306 in FIG. 3. Correspondingly, values of the similarity measure (220 in FIG. 2) that fall within the time interval 306 may be used to determine the LTP portion of the plasticity rule. In some implementations, the similarity measure and w(Δt) of plasticity adjustments may be scaled such that the mean value is constrained to be 1, or the maximum value may be constrained to be 1. The LTD portion of the STDP rule may comprise for example a monotonic function of time (e.g., a constant 302_1, gradually increasing and/or decreasing 302_2). The magnitude of the LTD portion may be determined based on a statistical parameter of the LTP portion. The statistical parameter may comprise for instance mean, median, a percentile, maximum, etc.
The LTD portion may be characterized by a time window Td 308. In some implementations, the LTP and LTD windows 306, 308 may be configured equal to one another. In one or more implementations, the time window ΔT used for determining similarity measure (e.g., the measure 220 in FIG. 2) may be set equal to either Tp, Td. The time windows ΔT, Tp, Td may also be configured different from one another, such that the window of LTD is longer than LTP (or vice versa), depending on the temporal structure of the pattern to be learned.
The adaptively configured STDP rules (e.g., the rules 304, 302 in FIG. 3) may be used to adjust efficacy of neuron connections. In some implementations, the adjustment may be performed based on the post-synaptic response of the neuron. In one or more implementations, the efficacy may comprise synaptic weight.
Returning now to FIG. 2, when the neuron generates post-synaptic response (230 in FIG. 2), the history of pre-synaptic input may be evaluated. One or more pre-synaptic pulses (the spikes 232, 234, 236 in FIG. 2) may be identified within the time window ΔT, prior to the post-synaptic spike 230. Plasticity components (e.g., the components 242, 244, 246), corresponding to the pre-synaptic times tpre i of the identified pulses 232, 234, 236, may be combined by an operator 240 to produce weight adjustment w 248. In some implementations, the adjustment may be performed based on an event, such as timer expiration, buffer overflow, external (reward) indication, and/or other event types).
Referring now to FIGS. 4-6, exemplary implementations of adaptive plasticity methods according to the disclosure are described. In some implementations, the methods of FIGS. 4-6 may be used, for example, for operating the neurons 102 of FIG. 1. Moreover, methods of FIG. 4-6 may be implemented in a connection (e.g., the connection 104, 124 of FIG. 1). The methods of FIG. 4-6 may also be implemented in sensory processing apparatus, comprising one or more spiking neuron networks as described with respect to FIG. 8, infra, thereby advantageously aiding, inter alia improving the speed and or quality of learning, potentially requiring fewer neurons to represent a given input, as well as learning temporally stable patterns.
FIG. 4 illustrates a method of determining adaptive plasticity for a spiking neuron based on a similarity measure between neuron input and output, in accordance with one implementation.
At step 402 of method 400, a neuron generates an output Y. In some implementations, the output may be generated in accordance with a deterministic spike response model, as described for example in co-owned U.S. patent application Ser. No. 13/152,119, entitled “APPARATUS AND METHODS FOR TEMPORALLY PROXIMATE OBJECT RECOGNITION”, filed on Jun. 2, 2011, and patented as U.S. Pat. No. 8,942,466 on Jan. 27, 2015, incorporated herein by reference in its entirety. In some implementations, the spike response process may comprise adaptive threshold adjustment as described in co-owned U.S. patent application Ser. No. 13/623,820, entitled “APPARATUS AND METHODS FOR ENCODING OF SENSORY DATA USING ARTIFICIAL SPIKING NEURONS”, filed on Sep. 20, 2012, incorporated herein by reference in its entirety. In some implementations, the neuron may be operable in accordance with stochastic process, as described in co-owned U.S. patent application Ser. No. 13/487,499, entitled “STOCHASTIC APPARATUS AND METHODS FOR IMPLEMENTING GENERALIZED LEARNING RULES”, filed on Jun. 4, 2012, incorporated herein by reference in its entirety.
At step 404 of the method 400, a history of inputs into the neuron may be accessed. In some implementations, the history may comprise one or more spikes 106_1, 106_1, 106_1 in FIG. 1. The history may comprise for example time data of pre-synaptic spikes stored in a synaptic memory buffer as described in U.S. patent application Ser. No. 13/239,259, entitled “APPARATUS AND METHOD FOR PARTIAL EVALUATION OF SYNAPTIC UPDATES BASED ON SYSTEM EVENTS”, filed on Sep. 21, 2011, and patented as U.S. Pat. No. 8,725,662 on May 13, 2014, incorporated supra.
At step 406, a similarity measure C(X,Y,t) between the output (X) and the input (Y) as a function of time shift (t) may be determined. In some implementations, the similarity measure may use correlation measured at different time shifts; i.e., a correlogram, as described with respect to FIG. 2, supra. The similarity measure may use for example mutual information at different time shifts as determined as follows:
I ( X ; Y ) = y Y x X p ( x , y ) log ( p ( x , y ) p ( x ) p ( y ) ) ,
where:
p(x,y) is the joint probability distribution function of X and Y; and
p(x) and p(y) are the marginal probability distribution functions of X and Y respectively.
At step 408, a plasticity rule is configured based on the similarity measure. In some implementations, the plasticity rule determination may comprise approach described with respect to FIGS. 2A-3, supra.
FIG. 5 illustrates a method of determining adaptive plasticity for a spiking neuron comprising multiple input connections, in accordance with one implementation.
At step 502 of the method 500, plasticity time window is configured. In some implementations, the LTP and LTD component may comprise the time extent T. In one or more implementations, the LTP and LTD component may comprise different time extents Tp, Td.
At step 504, LTD time-dependence is configured. In one or more implementations, the LTD may comprise a constant, a gradually increasing and/or decreasing function (e.g., the functions 302_1, 302_2 illustrated in FIG. 3).
At step 506, an event is detected. In some implementations, the event may comprise post-synaptic spike generation by the neuron. The event may comprise for example an external event (e.g., reinforcement signal), a timer event (e.g., for cyclic updates), a buffer overflow event (e.g., indicative of a memory buffer, storing, for example, pre-synaptic and/or post-synaptic spike history being full or nearly full), etc.
Responsive to the event, at step 508, a correlogram Ci (e.g., between pre-synaptic spike train xi from ith connection and the post-synaptic spike train Y over the time window T) is determined. In one implementation, the correlogram determination comprises:
    • partitioning the time window into M bins;
    • identifying number nm of pre-synaptic inputs xi associated with individual mth bin, m=1−M; and
    • incrementing value of mth bin by nm.
It will be appreciated by those skilled in the arts that other similarity measures may be employed at step 508 in place and/or along with the correlogram, such as, for example, cross-correlation, mutual information, and/or convolution.
At step 510, a check may be performed if inputs form other connections need to be processed. When data from additional connections exist, the method 500 may return to step 508.
When no additional data are present, the method 500 may proceed to step 512, where average correlogram may be determined as a weighted average:
<C>=ρ i=1 N(a i C i)  (Eqn. 1)
where ai comprise the weights. The individual weights may be set to the same value. In some implementations, the weights may be set synapse-specific. For example, the weights could be “0” for very weak synaptic connections, and “1” for strong connections.
At step 514, the LTP portion of the STDP rule (e.g., the rule 304 in FIG. 3) is determined using the causal (tpre<tpost) portion of the average correlogram <C> of Eqn. 1.
At step 516, the magnitude of the LTD portion of the STDP rule (e.g., the magnitude 310 of the rule 302_1 in FIG. 3) is determined. In some implementations, the magnitude may be based on a statistical parameter (e.g., mean, median, percentile, maximum, etc. of the LTP rule portion).
FIG. 6 illustrates a method of connection plasticity update based on the adaptive STDP rule, in accordance with one implementation.
At step 602, a determination may be made whether the update is to be performed. In one or more implementations, the update may be based on a post-synaptic spike by the neuron. The update may be based for example on an external event (e.g., reinforcement signal), a timer event (e.g., for cyclic updates), a buffer overflow event (e.g., indicative of a memory buffer, storing, for example, pre-synaptic and/or post-synaptic spike history being full or nearly full), etc.
When the update is to be performed, the method proceeds to step 604, where pre-synaptic update may be performed. In some implementations, the pre-synaptic update may comprise the LTD rule (e.g., the rule determined at step 516 of FIG. 5 and/or rule 302 of FIG. 3).
At step 606, the LTP and the LTD rules (e.g., the rules 304, 302 of FIG. 3) may or may not be combined to produce the post-rule. For example, in some implementations, the post-rule may be exclusively LTP-based, or may be a combination of both LTP and LTD. The combination may be used in order increase the selectivity of which temporal patterns are potentiated.
At step 608, the post-synaptic update is performed. In some implementations, the post-synaptic update may comprise the combined rule.
Performance
FIG. 7 illustrates exemplary performance data associated with a learning object recognition apparatus comprising an adaptive plasticity mechanism. A signal processing apparatus (e.g., the apparatus 1000 of FIG. 8 described in detail below) may receive one or more input frames, such as the frame 700 of FIG. 7. The input frames may comprise representations of one or more objects, such as the triangles 702 and the squares 704 of frame 700.
The circular panels in FIG. 7 depict receptive fields of the post-synaptic neurons of the network that may develop during learning by the network based on the input (e.g., one or more frames 700). The solid dots 718 in FIG. denote responses of the one or more post-synaptic neurons.
The receptive fields 710, 714, 716 depict performance of a network comprising plasticity mechanism of the prior art. As shown by the receptive fields 710, 714, 716, the network is capable of learning only a single feature (e.g., the triangle 702). Furthermore, the prior art plasticity mechanism may cause one or more post-synaptic neurons to generate erroneous responses, as illustrated by the solid circle 718 in FIG. 7.
Contrast the network responses 710, 714, 716 with the receptive 720, 722, 724, 726 corresponding to the network comprising an exemplary adaptive plasticity mechanism of the present disclosure. The receptive fields 720, 722, 724, 726 illustrate that the network (e.g., comprising one or more neurons 110 of FIG. 1) is capable of learning both the triangular objects (as shown by the neuron responses in the panels 720, 722) as well as the square objects (as shown by the neuron responses in the panels 724, 726). In addition, the performance of the network comprising the exemplary inventive adaptive plasticity mechanism is characterized by a lower number of erroneous responses.
Exemplary Apparatus
Various exemplary spiking network apparatus implementing one or more of the methods set forth herein (e.g., using the exemplary adaptive mechanisms described above) are now described with respect to FIGS. 7-11.
One apparatus for processing of sensory information (e.g., visual, audio, somatosensory) using a spiking neural network (including one or more of the conditional plasticity mechanisms described herein) is shown in FIG. 8. The illustrated processing apparatus 1000 includes an input interface configured to receive an input sensory signal 1020. In some implementations, this sensory input comprises electromagnetic waves (e.g., visible light, IR, UV, etc.) entering an imaging sensor array (comprising RGCs, a charge coupled device (CCD), CMOS device, or an active-pixel sensor (APS)). The input signal in this example is a sequence of images (image frames) received from a CCD or CMOS camera via a receiver apparatus, or downloaded from a file. Alternatively, the image may be a two-dimensional matrix of RGB values refreshed at a 24 Hz frame rate. It will be appreciated by those skilled in the art that the above image parameters and components are merely exemplary, and many other image representations (e.g., bitmap, CMYK, grayscale, etc.) and/or frame rates are equally useful with the present disclosure.
The apparatus 1000 may also include an encoder 1024 configured to transform (encode) the input signal so as to form an encoded signal 1026. In one variant, the encoded signal comprises a plurality of pulses (also referred to as a group of pulses) configured to model neuron behavior. The encoded signal 1026 may be communicated from the encoder 1024 via multiple connections (also referred to as transmission channels, communication channels, or synaptic connections) 1004 to one or more neuronal nodes (also referred to as the detectors) 1002.
In the implementation of FIG. 8, different detectors of the same hierarchical layer are denoted by an “_n” designator, such that e.g., the designator 1002_1 denotes the first detector of the layer 1002. Although only two detectors (1002_1, 1002 n) are shown in FIG. 8 for clarity, it is appreciated that the encoder can be coupled to any number of detector nodes that is compatible with the detection apparatus hardware and software limitations. Furthermore, a single detector node may be coupled to any practical number of encoders.
In one implementation, each of the detectors 1002_1, 1002 n contain logic (which may be implemented as a software code, hardware logic, or a combination of thereof) configured to recognize a predetermined pattern of pulses in the encoded signal 1004, using for example any of the mechanisms described in co-owned U.S. patent application Ser. No. 12/869,573, filed Aug. 26, 2010, entitled “SYSTEMS AND METHODS FOR INVARIANT PULSE LATENCY CODING”, and patent as U.S. Pat. No. 8,315,305 on Nov. 20, 2012, co-owned U.S. patent application Ser. No. 12/869,583, filed Aug. 26, 2010, entitled “INVARIANT PULSE LATENCY CODING SYSTEMS AND METHODS”, and patented as U.S. Pat. No. 8,467,623 on Jun. 18, 2013, co-owned and co-pending U.S. patent application Ser. No. 13/117,048, filed May 26, 2011 and entitled “APPARATUS AND METHODS FOR POLYCHRONOUS ENCODING AND MULTIPLEXING IN NEURONAL PROSTHETIC DEVICES”, co-owned and co-pending U.S. patent application Ser. No. 13/152,084, filed Jun. 2, 2011, entitled “APPARATUS AND METHODS FOR PULSE-CODE INVARIANT OBJECT RECOGNITION”, each incorporated herein by reference in its entirety, to produce post-synaptic detection signals transmitted over communication channels 1008. In FIG. 8, the designators 1008_1, 1008 n denote output of the detectors 1002_1, 1002 n, respectively.
In one implementation, the detection signals are delivered to a next layer of the detectors 1012 (comprising detectors 1012 l, 1012 m, 1012 k) for recognition of complex object features and objects, similar to the exemplary configuration described in commonly owned and co-pending U.S. patent application Ser. No. 13/152,084, filed Jun. 2, 2011, entitled “APPARATUS AND METHODS FOR PULSE-CODE INVARIANT OBJECT RECOGNITION”, incorporated herein by reference in its entirety. In this configuration, each subsequent layer of detectors is configured to receive signals from the previous detector layer, and to detect more complex features and objects (as compared to the features detected by the preceding detector layer). For example, a bank of edge detectors is followed by a bank of bar detectors, followed by a bank of corner detectors, and so on, thereby enabling alphabet recognition by the apparatus.
Each of the detectors 1002 may output detection (post-synaptic) signals on communication channels 1008_1, 1008 n (with appropriate latency) that may propagate with different conduction delays to the detectors 1012. The detector cascade of the apparatus of FIG. 8 may contain any practical number of detector nodes and detector banks determined, inter alia, by the software/hardware resources of the detection apparatus and complexity of the objects being detected.
The sensory processing apparatus implementation illustrated in FIG. 8 may further comprise lateral connections 1006. In some variants, the connections 1006 are configured to communicate post-synaptic activity indications between neighboring neurons of the same hierarchy level, as illustrated by the connection 1006_1 in FIG. 8. In some variants, the neighboring neuron may comprise neurons having overlapping inputs (e.g., the inputs 1004_1, 1004 n in FIG. 8), so that the neurons may compete in order to not learn the same input features. In one or more implementations, the neighboring neurons may comprise spatially proximate neurons such as being disposed within a certain volume/area from one another on a 3-dimensional (3D) and or two-dimensional (2D) space.
The apparatus 1000 may also comprise feedback connections 1014, configured to communicate context information from detectors within one hierarchy layer to previous layers, as illustrated by the feedback connections 1014_1 in FIG. 8. In some implementations, the feedback connection 1014_2 is configured to provide feedback to the encoder 1024 thereby facilitating sensory input encoding, as described in detail in commonly owned and co-pending U.S. patent application Ser. No. 13/152,084, filed Jun. 2, 2011, entitled “APPARATUS AND METHODS FOR PULSE-CODE INVARIANT OBJECT RECOGNITION”, incorporated supra.
One particular implementation of the computerized neuromorphic processing system, adapted for operating a computerized spiking network (and implementing the exemplary conditional plasticity methodology described supra), is illustrated in FIG. 9. The computerized system 1100 of FIG. 9 comprises an input interface 1110, such as for example an image sensor, a computerized spiking retina, an audio array, a touch-sensitive input device, etc. The input interface 1110 is coupled to the processing block (e.g., a single or multi-processor block) via the input communication interface 1114. The system 1100 further comprises a random access memory (RAM) 1108, configured to store neuronal states and connection parameters (e.g., weights 108 in FIG. 1), and to facilitate synaptic updates. In some exemplary implementations, synaptic updates are performed according to the description provided in, for example, in U.S. patent application Ser. No. 13/239,255 filed Sep. 21, 2011, entitled “APPARATUS AND METHODS FOR SYNAPTIC UPDATE IN A PULSE-CODED NETWORK”, incorporated by reference supra.
In some implementations, the memory 1108 is coupled to the processor 1102 via a direct connection (memory bus) 1116. The memory 1108 may also be coupled to the processor 1102 via a high-speed processor bus 1112).
The system 1100 may further comprise a nonvolatile storage device 1106, comprising, inter alia, computer readable instructions configured to implement various aspects of spiking neuronal network operation (e.g., sensory input encoding, connection plasticity, operation model of neurons, etc.). The nonvolatile storage 1106 may be used for instance to store state information of the neurons and connections when, for example, saving/loading network state snapshot, or implementing context switching (e.g., saving current network configuration (comprising, inter alia, connection weights and update rules, neuronal states and learning rules, etc.) for later use, and loading of a previously stored network configuration.
In some implementations, the computerized apparatus 1100 is coupled to one or more external processing/storage/input devices via an I/O interface 1120, such as a computer I/O bus (PCI-E), wired (e.g., Ethernet) or wireless (e.g., Wi-Fi) network connection.
In another variant, the input/output interface comprises a speech input (e.g., a microphone) and a speech recognition module configured to receive and recognize user commands.
It will be appreciated by those skilled in the arts that various processing devices may be used with computerized system 1100, including but not limited to, a single core/multicore CPU, DSP, FPGA, GPU, ASIC, combinations thereof, and/or other processors. Various user input/output interfaces are similarly applicable to implementations of the disclosure including, for example, an LCD/LED monitor, touch-screen input and display device, speech input device, stylus, light pen, trackball, end the likes.
Referring now to FIG. 10, one implementation of neuromorphic computerized system configured to implement a conditional plasticity mechanism in a spiking network is described in detail. The neuromorphic processing system 1130 of FIG. 10 comprises a plurality of processing blocks (micro-blocks) 1140, where each micro-block comprises a computing logic core 1132 and a memory block 1134. The logic core 1132 is configured to implement various aspects of neuronal node operation, such as the node model, and synaptic update rules (e.g., the I-STDP) and/or other tasks relevant to network operation. The memory block is configured to store, inter alia, neuronal state variables and connection parameters (e.g., weights, delays, I/O mapping) of connections 1138.
The micro-blocks 1140 are interconnected with one another using connections 1138 and routers 1136. As it is appreciated by those skilled in the arts, the connection layout in FIG. 10 is exemplary, and many other connection implementations (e.g., one to all, all to all, etc.) are compatible with the disclosure.
The neuromorphic apparatus 1130 is configured to receive input (e.g., visual input) via the interface 1142. In one or more implementations, applicable for example to interfacing with a computerized spiking retina or an image array, the apparatus 1130 may provide feedback information via the interface 1142 to facilitate encoding of the input signal.
The neuromorphic apparatus 1130 is configured to provide output (e.g., an indication of recognized object or a feature, or a motor command, e.g., to zoom/pan the image array) via the interface 1144.
The apparatus 1130, in one or more implementations, may interface to external fast response memory (e.g., RAM) via high bandwidth memory interface 1148, thereby enabling storage of intermediate network operational parameters (e.g., spike timing, etc.). The apparatus 1130 may also interface to external slower memory (e.g., Flash, or magnetic (hard drive)) via lower bandwidth memory interface 1146, in order to facilitate program loading, operational mode changes, and retargeting, where network node and connection information for a current task is saved for future use and flushed, and previously stored network configuration is loaded in its place.
FIG. 11 A illustrates implementations of a shared bus neuromorphic computerized system comprising micro-blocks 1140, described with respect to FIG. 10, supra, coupled to a shared interconnect. The apparatus 1145 of FIG. 11A utilizes one (or more) shared bus(es) 1146 in order to interconnect micro-blocks 1140 with one another.
FIG. 11B illustrates one implementation of cell-based neuromorphic computerized system architecture configured to implement Conditional plasticity mechanism in a spiking network. The neuromorphic system 1150 of FIG. 11B comprises a hierarchy of processing blocks (cells block). In some implementations, the lowest level L1 cell 1152 of the apparatus 1150 may comprise logic and memory, and may be configured similar to the micro block 1140 of the apparatus shown in FIG. 10. A number of cell blocks may be arranges in a cluster and communicate with one another a local interconnects 1162, 1164. Each such cluster may form a higher-level cell, e.g., cell L2, denoted as 1154 in FIG. 11B. Similarly, several L2 clusters may communicate with one another via a second-level interconnect 1166 and form a super-cluster L3, denoted as 1156 in FIG. 11B. The super-clusters 1154 may for example communicate via a third level interconnect 1168, and may form a next level cluster, and so on. It will be appreciated by those skilled in the arts that the hierarchical structure of the apparatus 1150, comprising a given number (e.g., four) cells per level, is merely one exemplary implementation, and other implementations may comprise more or fewer cells per level, and/or fewer or more levels, as well as yet other types of architectures.
Different cell levels (e.g., L1, L2, L3) of the exemplary apparatus 1150 of FIG. 11B may be configured to perform functionality with various levels of complexity. In one implementation, different L1 cells may process in parallel different portions of the visual input (e.g., encode different frame macro-blocks), with the L2, L3 cells performing progressively higher-level functionality (e.g., edge detection, object detection). Different L2, L3 cells may also perform different aspects of operating for example a robot, with one or more L2/L3 cells processing visual data from a camera, and other L2/L3 cells operating a motor control block for implementing lens motion when e.g., tracking an object, or performing lens stabilization functions.
The neuromorphic apparatus 1150 may receive input (e.g., visual input) via the interface 1160. In one or more implementations, applicable for example to interfacing with a computerized spiking retina or image array, the apparatus 1150 may provide feedback information via the interface 1160 to facilitate encoding of the input signal.
The neuromorphic apparatus 1150 may provide output (e.g., an indication of recognized object or a feature, or a motor command, e.g., to zoom/pan the image array) via the interface 1170. In some implementations, the apparatus 1150 may perform all of the I/O functionality using single I/O block (not shown).
The apparatus 1150, in one or more implementations, may also interface to external fast response memory (e.g., RAM) via high bandwidth memory interface (not shown), thereby enabling storage of intermediate network operational parameters (e.g., spike timing, etc.). The apparatus 1150 may also interface to external slower memory (e.g., flash, or magnetic (hard drive)) via lower bandwidth memory interface (not shown), in order to facilitate program loading, operational mode changes, and retargeting, where network node and connection information for a current task is saved for future use and flushed, and a previously stored network configuration is loaded in its place.
Exemplary Uses and Applications of Certain Aspects of the Disclosure
The plasticity mechanism described herein may advantageously improve the quality of the feature learning (as characterized by a number of false positives) and/or to increase learning speed. The adaptive plasticity mechanism may enable the processing apparatus to learn temporally stable patterns.
Various aspects of the disclosure may advantageously be applied to, inter alia, the design and operation of large spiking neural networks configured to process streams of input stimuli, in order to aid in detection and functional binding related aspect of the input.
In some implementations, conditional plasticity mechanisms described herein may be implemented in a spiking neuron of a network, or in a connection of the network.
The approach of the disclosure can advantageously, among other things, achieve faster learning and improve the quality of learned features and (iv) enable efficient network operation (e.g., visual input encoding) for a wide variety of input conditions. It will be appreciated that the increased network stability and flexibility may be traded for (a) a less complex, less costly and more robust network capable of processing the same feature set with fewer neurons; and/or (b) a more capable, higher performance network capable of processing larger and more complex feature set with the same number of neurons, when compared to the prior art solutions.
It is appreciated by those skilled in the arts that above implementation are exemplary, and the framework of the disclosure is equally compatible and applicable to processing of other information, such as, for example information classification using a database, where the detection of a particular pattern can be identified as a discrete signal similar to a spike, and where coincident detection of other patterns influences detection of a particular one pattern based on a history of previous detections in a way similar to an operation of exemplary spiking neural network.
Advantageously, exemplary implementations of the various aspects of the present innovation are useful in a variety of devices including without limitation prosthetic devices, autonomous and robotic apparatus, and other electromechanical devices requiring sensory processing functionality. Examples of such robotic devises are manufacturing robots (e.g., automotive), military, medical (e.g. processing of microscopy, x-ray, ultrasonography, tomography). Examples of autonomous vehicles include rovers, unmanned air vehicles, underwater vehicles, smart appliances (e.g. ROOMBA®), etc.
Implementations of the principles of the disclosure are applicable to video data compression and processing in a wide variety of stationary and portable devices, such as, for example, smart phones, portable communication devices, notebook, netbook and tablet computers, surveillance camera systems, and practically any other computerized device configured to process vision data
Implementations of the principles of the disclosure are further applicable to a wide assortment of applications including computer human interaction (e.g., recognition of gestures, voice, posture, face, etc.), controlling processes (e.g., an industrial robot, autonomous and other vehicles), augmented reality applications, organization of information (e.g., for indexing databases of images and image sequences), access control (e.g., opening a door based on a gesture, opening an access way based on detection of an authorized person), detecting events (e.g., for visual surveillance or people or animal counting, tracking), data input, financial transactions (payment processing based on recognition of a person or a special payment symbol) and many others.
Advantageously, the disclosure can be used to simplify tasks related to motion estimation, such as where an image sequence is processed to produce an estimate of the object position (and hence velocity) either at each points in the image or in the 3D scene, or even of the camera that produces the images. Examples of such tasks are: ego motion, i.e., determining the three-dimensional rigid motion (rotation and translation) of the camera from an image sequence produced by the camera; following the movements of a set of interest points or objects (e.g., vehicles or humans) in the image sequence and with respect to the image plane.
In another approach, portions of the object recognition system are embodied in a remote server, comprising a computer readable apparatus storing computer executable instructions configured to perform pattern recognition in data streams for various applications, such as scientific, geophysical exploration, surveillance, navigation, data mining (e.g., content-based image retrieval). Myriad other applications exist that will be recognized by those of ordinary skill given the present disclosure.

Claims (24)

What is claimed:
1. Computerized spiking neuron apparatus comprising a non-transitory computer readable storage medium, the storage medium comprising a plurality of executable instructions configured to, adjust an efficacy of an interface of the neuron, the plurality of executable instructions are configured to. when executed, cause the computerized spiking neuron apparatus to:
adjust a neuron excitability based on one or more pre-synaptic inputs;
store one or more pre-synaptic times associated with the one or more pre-synaptic inputs; and
when the neuron excitability is above a threshold:
generate a post-synaptic output;
determine a similarity measure between a neuron output signal, comprising a response, and the one or more pre-synaptic inputs;
determine a long term potentiation (LTP) component of an adaptive plasticity based on the determined similarity measure; and
determine a long term depression (LTD) component of the adaptive plasticity based on a parameter associated with the LTP;
wherein the post-synaptic output comprises a response generated by the neuron based on the one or more pre-synaptic inputs.
2. The computerized spiking neuron apparatus of claim 1, wherein:
the similarity measure comprises a histogram characterized by a plurality of time interval bins;
a bin count associated with the histogram is determined based on individual ones of the one or more pre-synaptic times falling within respective ones of the plurality of time interval bins; and
a magnitude associated with the histogram is scaled according to any of (i) a constant mean value as a function of time; or (ii) a constant maximum value as a function of time.
3. The computerized spiking neuron apparatus of claim 1, wherein:
the one or more pre-synaptic inputs comprise one or more spike-trains provided to the neuron via one or more respective connections; and
the similarity measure is determined based on one or more correlograms, individual ones of the one or more correlograms being based on a correlation between the post-synaptic output and individual ones of the one or more spike-trains.
4. The computerized spiking neuron apparatus of claim 3, wherein:
the one or more correlograms comprise at least two correlograms; and
the similarity measure is determined based on a weighted average of the at least two correlograms, a combination characterized by a plurality of weights, individual ones of the plurality of weights being determined in accordance with weights of individual ones of the one or more connections.
5. A computerized spiking neuron network system configured to implement an adaptive plasticity in the network, the system comprising:
one or more processors configured to execute computer program modules, the execution of the computer program modules configured to cause the one or more processors, based on a response by a neuron of the network, to adjust one or more connections capable of providing input to the neuron by at least, where the execution of the computer program modules cause the one or more processors to:
determine one or more similarity measures between a neuron output signal, comprising a response, and one or more input signals provided by respective ones of the one or more connections;
determine a long term potentiation (LTP) component of an adaptive plasticity based on an average of the determined one or more similarity measures; and
determine a long term depression (LTD) component of the adaptive plasticity based on a parameter associated with the LTP.
6. A method of updating an efficacy of a connection configured to communicate an input to an artificial spiking neuron, the method comprising:
determining a similarity measure between the input and a neuron output;
identifying one or more components associated with one or more lag values, respectively, from the similarity measure; and
updating the efficacy based on the identified one or more components;
wherein:
the similarity measure is determined based on any of (i) a cross-correlation between the input and the neuron output; (ii) a correlogram between the input and the neuron output; and/or (iii) mutual information between the input and the neuron output;
the neuron output comprises one or more output spikes;
for individual ones of the one or more output spikes, the similarity measure is determined based on a convolution of the neuron output and at least a portion of the input, the input comprising one or more input spikes characterized by one or more pre-synaptic times, the at least portion corresponding to a window preceding the individual ones of the one or more output spikes; and
the similarity measure is characterized by the one or more lag values based on one or more time intervals between times of the individual ones of the one or more output spikes and respective ones of the one or more pre-synaptic times.
7. The method of claim 6, wherein:
a plurality of similarity measures is determined for a plurality of the one or more output spikes; and
the similarity measure is determined based on an average of individual ones of the plurality of similarity measures.
8. The method of claim 7, wherein a number of spikes within the plurality of the one or more output spikes is selected within a time interval ranging between 1 and 100 ms.
9. The method of claim 7, wherein a number of spikes within the plurality of the one or more output spikes is determined based on an averaging interval selected within the range between 100 spikes and 100,000 spikes.
10. The method of claim 6, wherein a connection plasticity update is based on an event selected from the group comprising: (i) a timer expiration; (ii) an overflow of a buffer, the buffer being capable of storing one or more spike times associated with the input; and
(iii) an external trigger.
11. The method of claim 10, wherein the event comprises the external trigger, and the external trigger is provided to the artificial spiking neuron via a bus separate from the connection.
12. The method of claim 10, wherein a combination of the identified one or more components comprises a weighted sum of individual ones of components associated with at least one spike of the input occurring within an interval associated with the one or more lag values.
13. The method of claim 6, wherein:
the similarity measure comprises a potentiation portion configured to potentiate the connection, and a depression portion configured to depress the connection;
the potentiation of the connection comprises substantially increasing the efficacy thereof; and
the depression of the connection comprises substantially reducing the efficacy thereof.
14. The method of claim 13, wherein:
the efficacy is characterized by a connection weight;
the potentiation of the connection is characterized by a first time-dependent function having a first time window associated therewith; and
the potentiation of the connection comprises substantially increasing the connection weight, the increase of the connection weight being based at least on an integral of the first time-dependent function over the first time window, the first time window being a positive value.
15. The method of claim 14, wherein:
the depression of the connection is characterized by a second time-dependent function having a second time window associated therewith; and
the depression of the connection comprises substantially decreasing the connection weight, the decrease of the connection weight being based at least on an integral of the second time-dependent function over the second time window, the second time window being a negative value.
16. The method of claim 14, wherein:
the input is configured to convey information associated with an image element; and
the positive value is configured to increase the connection weight, causing an increase of a probability of generating the neuron output based on the artificial spiking neuron identifying the image element.
17. The method of claim 15, wherein at least one of the first and second time windows are configured to be between 5 ms and 50 ms inclusive.
18. The method of claim 13, wherein:
the efficacy is characterized by a probability of transmission;
the potentiation of the connection is characterized by a first time-dependent function having a first time window associated therewith; and
the potentiation of the connection comprises substantially increasing the probability of transmission, the increase of the probability of transmission being based at least on an integral of the first time-dependent function over the first time window, the first time window being a positive value.
19. The method of claim 6, wherein:
the identified one or more components comprise at least two components; and the updating the efficacy is further based on a combination of the at least two components.
20. The method of claim 7, wherein a number of spikes within the plurality of the one or more output spikes is determined based on an averaging interval selected within the range between 10 seconds and 10,000 seconds.
21. A method of updating an efficacy of a connection configured to communicate an input to an artificial spiking neuron, the method comprising:
determining a similarity measure between the input and a neuron output;
identifying one or more components associated with one or more lag values, respectively, from the similarity measure; and
updating the efficacy based on the identified one or more components;
wherein:
the similarity measure comprises a potentiation portion configured to potentiate the connection, and a depression portion configured to depress the connection;
the potentiation of the connection comprises substantially increasing the efficacy thereof;
the depression of the connection comprises substantially reducing the efficacy thereof;
the efficacy is characterized by a connection weight;
the potentiation of the connection is characterized by a first time-dependent function having a first time window associated therewith; and
the potentiation of the connection comprises substantially increasing the connection weight, the increase of the connection weight being based at least on an integral of the first time-dependent function over the first time window, the first time window being a positive value.
22. The method of claim 21, wherein:
the depression of the connection is characterized by a second time-dependent function having a second time window associated therewith; and
the depression of the connection comprises substantially decreasing the connection weight, the decrease of the connection weight based on an integral of the second time-dependent function over the second time window, the second time window being a negative value.
23. A method of updating an efficacy of a connection configured to communicate an input to an artificial spiking neuron, the method comprising:
determining a similarity measure between the input and a neuron output;
identifying one or more components associated with one or more lag values, respectively, from the similarity measure; and
updating the efficacy based on the identified one or more components;
wherein:
the similarity measure comprises a potentiation portion configured to potentiate the connection, and a depression portion configured to depress the connection;
the potentiation of the connection comprises substantially increasing the efficacy thereof;
the depression of the connection comprises substantially reducing the efficacy thereof;
the efficacy is characterized by a probability of transmission;
the potentiation of the connection is characterized by a first time-dependent function having a first time window associated therewith; and
the potentiation of the connection comprises substantially increasing the probability of transmission, the increase of the probability of transmission being based at least on an integral of the first time-dependent function over the first time window, the first time window being a positive value.
24. The method of claim 23, wherein the first time window is configured between 5 ms and 50 ms inclusive.
US13/660,923 2012-10-25 2012-10-25 Adaptive plasticity apparatus and methods for spiking neuron network Active 2033-12-29 US9183493B2 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US13/660,923 US9183493B2 (en) 2012-10-25 2012-10-25 Adaptive plasticity apparatus and methods for spiking neuron network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
US13/660,923 US9183493B2 (en) 2012-10-25 2012-10-25 Adaptive plasticity apparatus and methods for spiking neuron network

Publications (2)

Publication Number Publication Date
US20140122397A1 US20140122397A1 (en) 2014-05-01
US9183493B2 true US9183493B2 (en) 2015-11-10

Family

ID=50548324

Family Applications (1)

Application Number Title Priority Date Filing Date
US13/660,923 Active 2033-12-29 US9183493B2 (en) 2012-10-25 2012-10-25 Adaptive plasticity apparatus and methods for spiking neuron network

Country Status (1)

Country Link
US (1) US9183493B2 (en)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180173983A1 (en) * 2016-12-21 2018-06-21 Volkswagen Ag Digital neuromorphic (nm) sensor array, detector, engine and methodologies
US20180173992A1 (en) * 2016-12-21 2018-06-21 Volkswagen Ag Vector engine and methodologies using digital neuromorphic (nm) data
US10235565B2 (en) 2016-12-21 2019-03-19 Volkswagen Ag System and methodologies for occupant monitoring utilizing digital neuromorphic (NM) data and fovea tracking
US10282615B2 (en) 2016-12-21 2019-05-07 Volkswagen Ag System and method for root association in image data
US10789495B2 (en) 2016-12-21 2020-09-29 Volkswagen Ag System and method for 1D root association providing sparsity guarantee in image data
US10922824B1 (en) 2019-08-13 2021-02-16 Volkswagen Ag Object tracking using contour filters and scalers
US11080592B2 (en) 2017-10-05 2021-08-03 International Business Machines Corporation Neuromorphic architecture for feature learning using a spiking neural network
US11100396B2 (en) 2017-08-24 2021-08-24 International Business Machines Corporation Self-adjusting threshold for synaptic activity in neural networks
US11556343B2 (en) 2017-09-22 2023-01-17 International Business Machines Corporation Computational method for temporal pooling and correlation

Families Citing this family (61)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9311593B2 (en) 2010-03-26 2016-04-12 Brain Corporation Apparatus and methods for polychronous encoding and multiplexing in neuronal prosthetic devices
US8315305B2 (en) 2010-03-26 2012-11-20 Brain Corporation Systems and methods for invariant pulse latency coding
US9122994B2 (en) 2010-03-26 2015-09-01 Brain Corporation Apparatus and methods for temporally proximate object recognition
US9405975B2 (en) 2010-03-26 2016-08-02 Brain Corporation Apparatus and methods for pulse-code invariant object recognition
US8942466B2 (en) 2010-08-26 2015-01-27 Brain Corporation Sensory input processing apparatus and methods
US9193075B1 (en) 2010-08-26 2015-11-24 Brain Corporation Apparatus and methods for object detection via optical flow cancellation
US9015093B1 (en) 2010-10-26 2015-04-21 Michael Lamport Commons Intelligent control with hierarchical stacked neural networks
US8775341B1 (en) 2010-10-26 2014-07-08 Michael Lamport Commons Intelligent control with hierarchical stacked neural networks
US9566710B2 (en) 2011-06-02 2017-02-14 Brain Corporation Apparatus and methods for operating robotic devices using selective state space training
US9047568B1 (en) 2012-09-20 2015-06-02 Brain Corporation Apparatus and methods for encoding of sensory data using artificial spiking neurons
US9147156B2 (en) 2011-09-21 2015-09-29 Qualcomm Technologies Inc. Apparatus and methods for synaptic update in a pulse-coded network
US9070039B2 (en) 2013-02-01 2015-06-30 Brian Corporation Temporal winner takes all spiking neuron network sensory processing apparatus and methods
US20150074026A1 (en) * 2011-08-17 2015-03-12 Qualcomm Technologies Inc. Apparatus and methods for event-based plasticity in spiking neuron networks
US8725658B2 (en) 2011-09-21 2014-05-13 Brain Corporation Elementary network description for efficient memory management in neuromorphic systems
US8719199B2 (en) 2011-09-21 2014-05-06 Brain Corporation Systems and methods for providing a neural network having an elementary network description for efficient implementation of event-triggered plasticity rules
US8725662B2 (en) 2011-09-21 2014-05-13 Brain Corporation Apparatus and method for partial evaluation of synaptic updates based on system events
US9104973B2 (en) 2011-09-21 2015-08-11 Qualcomm Technologies Inc. Elementary network description for neuromorphic systems with plurality of doublets wherein doublet events rules are executed in parallel
US9117176B2 (en) 2011-09-21 2015-08-25 Qualcomm Technologies Inc. Round-trip engineering apparatus and methods for neural networks
US10210452B2 (en) 2011-09-21 2019-02-19 Qualcomm Incorporated High level neuromorphic network description apparatus and methods
US9098811B2 (en) 2012-06-04 2015-08-04 Brain Corporation Spiking neuron network apparatus and methods
US20130297539A1 (en) * 2012-05-07 2013-11-07 Filip Piekniewski Spiking neural network object recognition apparatus and methods
US9129221B2 (en) * 2012-05-07 2015-09-08 Brain Corporation Spiking neural network feedback apparatus and methods
US9224090B2 (en) * 2012-05-07 2015-12-29 Brain Corporation Sensory input processing apparatus in a spiking neural network
US9208432B2 (en) 2012-06-01 2015-12-08 Brain Corporation Neural network learning and collaboration apparatus and methods
US9412041B1 (en) 2012-06-29 2016-08-09 Brain Corporation Retinal apparatus and methods
US9111215B2 (en) 2012-07-03 2015-08-18 Brain Corporation Conditional plasticity spiking neuron network apparatus and methods
US8977582B2 (en) 2012-07-12 2015-03-10 Brain Corporation Spiking neuron network sensory processing apparatus and methods
US9311594B1 (en) 2012-09-20 2016-04-12 Brain Corporation Spiking neuron network apparatus and methods for encoding of sensory data
US9111226B2 (en) 2012-10-25 2015-08-18 Brain Corporation Modulated plasticity apparatus and methods for spiking neuron network
US9218563B2 (en) 2012-10-25 2015-12-22 Brain Corporation Spiking neuron sensory processing apparatus and methods for saliency detection
US9183493B2 (en) 2012-10-25 2015-11-10 Brain Corporation Adaptive plasticity apparatus and methods for spiking neuron network
US9275326B2 (en) 2012-11-30 2016-03-01 Brain Corporation Rate stabilization through plasticity in spiking neuron network
US9123127B2 (en) 2012-12-10 2015-09-01 Brain Corporation Contrast enhancement spiking neuron network sensory processing apparatus and methods
US9177245B2 (en) 2013-02-08 2015-11-03 Qualcomm Technologies Inc. Spiking network apparatus and method with bimodal spike-timing dependent plasticity
US9764468B2 (en) 2013-03-15 2017-09-19 Brain Corporation Adaptive predictor apparatus and methods
US9242372B2 (en) 2013-05-31 2016-01-26 Brain Corporation Adaptive robotic interface apparatus and methods
US9792546B2 (en) 2013-06-14 2017-10-17 Brain Corporation Hierarchical robotic controller apparatus and methods
US9314924B1 (en) 2013-06-14 2016-04-19 Brain Corporation Predictive robotic controller apparatus and methods
US9436909B2 (en) 2013-06-19 2016-09-06 Brain Corporation Increased dynamic range artificial neuron network apparatus and methods
US9239985B2 (en) 2013-06-19 2016-01-19 Brain Corporation Apparatus and methods for processing inputs in an artificial neuron network
US9552546B1 (en) 2013-07-30 2017-01-24 Brain Corporation Apparatus and methods for efficacy balancing in a spiking neuron network
US9579789B2 (en) 2013-09-27 2017-02-28 Brain Corporation Apparatus and methods for training of robotic control arbitration
US9489623B1 (en) 2013-10-15 2016-11-08 Brain Corporation Apparatus and methods for backward propagation of errors in a spiking neuron network
US9463571B2 (en) 2013-11-01 2016-10-11 Brian Corporation Apparatus and methods for online training of robots
US9597797B2 (en) 2013-11-01 2017-03-21 Brain Corporation Apparatus and methods for haptic training of robots
US9248569B2 (en) 2013-11-22 2016-02-02 Brain Corporation Discrepancy detection apparatus and methods for machine learning
US9358685B2 (en) 2014-02-03 2016-06-07 Brain Corporation Apparatus and methods for control of robot actions based on corrective user inputs
US9346167B2 (en) 2014-04-29 2016-05-24 Brain Corporation Trainable convolutional network apparatus and methods for operating a robotic vehicle
US9713982B2 (en) 2014-05-22 2017-07-25 Brain Corporation Apparatus and methods for robotic operation using video imagery
US9939253B2 (en) 2014-05-22 2018-04-10 Brain Corporation Apparatus and methods for distance estimation using multiple image sensors
US10194163B2 (en) 2014-05-22 2019-01-29 Brain Corporation Apparatus and methods for real time estimation of differential motion in live video
US9848112B2 (en) 2014-07-01 2017-12-19 Brain Corporation Optical detection apparatus and methods
US10057593B2 (en) 2014-07-08 2018-08-21 Brain Corporation Apparatus and methods for distance estimation using stereo imagery
US10055850B2 (en) 2014-09-19 2018-08-21 Brain Corporation Salient features tracking apparatus and methods using visual initialization
US9630318B2 (en) 2014-10-02 2017-04-25 Brain Corporation Feature detection apparatus and methods for training of robotic navigation
US9881349B1 (en) 2014-10-24 2018-01-30 Gopro, Inc. Apparatus and methods for computerized object identification
US9717387B1 (en) 2015-02-26 2017-08-01 Brain Corporation Apparatus and methods for programming and training of robotic household appliances
US10197664B2 (en) 2015-07-20 2019-02-05 Brain Corporation Apparatus and methods for detection of objects using broadband signals
US11238337B2 (en) * 2016-08-22 2022-02-01 Applied Brain Research Inc. Methods and systems for implementing dynamic neural networks
US11308387B2 (en) 2017-05-09 2022-04-19 Samsung Electronics Co., Ltd. STDP with synaptic fatigue for learning of spike-time-coded patterns in the presence of parallel rate-coding
US20200272883A1 (en) * 2017-12-19 2020-08-27 Intel Coporation Reward-based updating of synpatic weights with a spiking neural network

Citations (111)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5063603A (en) 1989-11-06 1991-11-05 David Sarnoff Research Center, Inc. Dynamic method for recognizing objects and image processing system therefor
US5138447A (en) 1991-02-11 1992-08-11 General Instrument Corporation Method and apparatus for communicating compressed digital video signals using multiple processors
US5216752A (en) 1990-12-19 1993-06-01 Baylor College Of Medicine Interspike interval decoding neural network
US5272535A (en) 1991-06-13 1993-12-21 Loral Fairchild Corporation Image sensor with exposure control, selectable interlaced, pseudo interlaced or non-interlaced readout and video compression
US5355435A (en) 1992-05-18 1994-10-11 New Mexico State University Technology Transfer Corp. Asynchronous temporal neural processing element
US5638359A (en) 1992-12-14 1997-06-10 Nokia Telecommunications Oy Method for congestion management in a frame relay network and a node in a frame relay network
US5652594A (en) 1967-12-28 1997-07-29 Lockheed Martin Corporation Signal processor affording improved immunity to medium anomalies and interference in remote object detection system
US5673367A (en) 1992-10-01 1997-09-30 Buckley; Theresa M. Method for neural network control of motion using real-time environmental feedback
RU2108612C1 (en) 1994-09-14 1998-04-10 Круглов Сергей Петрович Adaptive control system with identifier and implicit reference model
US5875108A (en) 1991-12-23 1999-02-23 Hoffberg; Steven M. Ergonomic man-machine interface incorporating adaptive pattern recognition based control system
US6009418A (en) 1996-05-02 1999-12-28 Cooper; David L. Method and apparatus for neural networking using semantic attractor architecture
US6014653A (en) 1996-01-26 2000-01-11 Thaler; Stephen L. Non-algorithmically implemented artificial neural networks and components thereof
US6035389A (en) 1998-08-11 2000-03-07 Intel Corporation Scheduling instructions with different latencies
US6418424B1 (en) 1991-12-23 2002-07-09 Steven M. Hoffberg Ergonomic man-machine interface incorporating adaptive pattern recognition based control system
US6458157B1 (en) 1997-08-04 2002-10-01 Suaning Gregg Joergen Retinal stimulator
US6509854B1 (en) 1997-03-16 2003-01-21 Hitachi, Ltd. DA conversion circuit
US20030050903A1 (en) 1997-06-11 2003-03-13 Jim-Shih Liaw Dynamic synapse for signal processing in neural networks
US6546291B2 (en) 2000-02-16 2003-04-08 Massachusetts Eye & Ear Infirmary Balance prosthesis
US6545705B1 (en) 1998-04-10 2003-04-08 Lynx System Developers, Inc. Camera with object recognition/data output
US6545708B1 (en) 1997-07-11 2003-04-08 Sony Corporation Camera controlling device and method for predicted viewing
US6581046B1 (en) 1997-10-10 2003-06-17 Yeda Research And Development Co. Ltd. Neuronal phase-locked loops
US6625317B1 (en) 1995-09-12 2003-09-23 Art Gaffin Visual imaging system and method
US20030216919A1 (en) 2002-05-13 2003-11-20 Roushar Joseph C. Multi-dimensional method and apparatus for automated language interpretation
US20040136439A1 (en) 2002-11-15 2004-07-15 Brandon Dewberry Methods and systems acquiring impulse signals
US20040170330A1 (en) 1998-08-12 2004-09-02 Pixonics, Inc. Video coding reconstruction apparatus and methods
US20040193670A1 (en) 2001-05-21 2004-09-30 Langan John D. Spatio-temporal filter and method
US20040220082A1 (en) * 2003-01-21 2004-11-04 Northwestern University Manipulation of neuronal ion channels
US20050015351A1 (en) 2003-07-18 2005-01-20 Alex Nugent Nanotechnology neural network methods and systems
US20050036649A1 (en) 2001-08-23 2005-02-17 Jun Yokono Robot apparatus, face recognition method, and face recognition apparatus
US20050283450A1 (en) 2004-06-11 2005-12-22 Masakazu Matsugu Information processing apparatus, information processing method, pattern recognition apparatus, and pattern recognition method
US20060094001A1 (en) 2002-11-29 2006-05-04 Torre Vicent E Method and device for image processing and learning with neuronal cultures
US7054850B2 (en) 2000-06-16 2006-05-30 Canon Kabushiki Kaisha Apparatus and method for detecting or recognizing pattern by employing a plurality of feature detecting elements
US20060129728A1 (en) 2004-12-09 2006-06-15 Hampel Craig E Memory interface with workload adaptive encode/decode
US20060161218A1 (en) 2003-11-26 2006-07-20 Wicab, Inc. Systems and methods for treating traumatic brain injury
US20070022068A1 (en) 2005-07-01 2007-01-25 Ralph Linsker Neural networks for prediction and control
US20070176643A1 (en) 2005-06-17 2007-08-02 Alex Nugent Universal logic gate utilizing nanotechnology
US20070208678A1 (en) 2004-03-17 2007-09-06 Canon Kabushiki Kaisha Parallel Pulse Signal Processing Apparatus, Pattern Recognition Apparatus, And Image Input Apparatus
US20080100482A1 (en) 2003-05-27 2008-05-01 Lazar Aurel A Multichannel Time Encoding And Decoding Of A Signal
JP4087423B2 (en) 2006-10-17 2008-05-21 京セラミタ株式会社 Portable communication device
WO2008083335A2 (en) 2006-12-29 2008-07-10 Neurosciences Research Foundation, Inc. Solving the distal reward problem through linkage of stdp and dopamine signaling
US20080199072A1 (en) 2003-02-27 2008-08-21 Sony Corporation Image processing device and method, learning device and method, recording medium, and program
US20080201282A1 (en) 2005-03-31 2008-08-21 France Telecom System and Method for Locating Points of Interest in an Object Image Implementing a Neural Network
US20080237446A1 (en) 2007-02-16 2008-10-02 Texas Instruments Incorporated Solid-state image pickup device and method
WO2008132066A1 (en) 2007-04-27 2008-11-06 Siemens Aktiengesellschaft A method for computer-assisted learning of one or more neural networks
US20090043722A1 (en) 2003-03-27 2009-02-12 Alex Nugent Adaptive neural network utilizing nanotechnology-based components
US7580907B1 (en) 2004-01-14 2009-08-25 Evolved Machines, Inc. Invariant object recognition
US7653255B2 (en) 2004-06-02 2010-01-26 Adobe Systems Incorporated Image region of interest encoding
US20100036457A1 (en) 2008-08-07 2010-02-11 Massachusetts Institute Of Technology Coding for visual prostheses
US20100081958A1 (en) 2006-10-02 2010-04-01 She Christy L Pulse-based feature extraction for neural recordings
US20100086171A1 (en) 2008-10-02 2010-04-08 Silverbrook Research Pty Ltd Method of imaging coding pattern having merged data symbols
US20100100482A1 (en) 2007-01-23 2010-04-22 Sxip Identity Corp. Intermediate Data Generation For Transaction Processing
US7737933B2 (en) 2000-09-26 2010-06-15 Toshiba Matsushita Display Technology Co., Ltd. Display unit and drive system thereof and an information display unit
US20100166320A1 (en) 2008-12-26 2010-07-01 Paquier Williams J F Multi-stage image pattern recognizer
US7765029B2 (en) 2005-09-13 2010-07-27 Neurosciences Research Foundation, Inc. Hybrid control device
US20100225824A1 (en) 2007-06-28 2010-09-09 The Trustees Of Columbia University In The City Of New York Multi-Input Multi-Output Time Encoding And Decoding Machines
US20100235310A1 (en) 2009-01-27 2010-09-16 Gage Fred H Temporally dynamic artificial neural networks
US20100299296A1 (en) 2009-05-21 2010-11-25 International Business Machines Corporation Electronic learning synapse with spike-timing dependent plasticity using unipolar memory-switching elements
US7849030B2 (en) 2006-05-31 2010-12-07 Hartford Fire Insurance Company Method and system for classifying documents
RU2406105C2 (en) 2006-06-13 2010-12-10 Филипп Геннадьевич Нестерук Method of processing information in neural networks
US20110016071A1 (en) 2009-07-20 2011-01-20 Guillen Marcos E Method for efficiently simulating the information processing in cells and tissues of the nervous system with a temporal series compressed encoding neural network
US20110119214A1 (en) 2009-11-18 2011-05-19 International Business Machines Corporation Area efficient neuromorphic circuits
US20110119215A1 (en) 2009-11-13 2011-05-19 International Business Machines Corporation Hardware analog-digital neural networks
US20110137843A1 (en) 2008-08-28 2011-06-09 Massachusetts Institute Of Technology Circuits and Methods Representative of Spike Timing Dependent Plasticity of Neurons
US20110160741A1 (en) 2008-06-09 2011-06-30 Hiroyuki Asano Medical treatment tool for tubular organ
RU2424561C2 (en) 2005-08-31 2011-07-20 Майкрософт Корпорейшн Training convolutional neural network on graphics processing units
US8000967B2 (en) 2005-03-09 2011-08-16 Telefonaktiebolaget Lm Ericsson (Publ) Low-complexity code excited linear prediction encoding
CN102226740A (en) 2011-04-18 2011-10-26 中国计量学院 Bearing fault detection method based on manner of controlling stochastic resonance by external periodic signal
US20120011090A1 (en) 2010-07-07 2012-01-12 Qualcomm Incorporated Methods and systems for three-memristor synapse with stdp and dopamine signaling
US20120083982A1 (en) 2010-10-05 2012-04-05 Zachary Thomas Bonefas System and method for governing a speed of an autonomous vehicle
US20120084240A1 (en) 2010-09-30 2012-04-05 International Business Machines Corporation Phase change memory synaptronic circuit for spiking computation, association and recall
US8154436B2 (en) 2005-10-24 2012-04-10 Mitsubishi Electric Information Technology Centre Europe B.V. Object detection
US20120109866A1 (en) 2010-10-29 2012-05-03 International Business Machines Corporation Compact cognitive synaptic computing circuits
US8281997B2 (en) 2008-02-19 2012-10-09 Bilcare Technologies Singapore Pte. Ltd. Reading device for identifying a tag or an object adapted to be identified, related methods and systems
US8315305B2 (en) 2010-03-26 2012-11-20 Brain Corporation Systems and methods for invariant pulse latency coding
US20120303091A1 (en) 2010-03-26 2012-11-29 Izhikevich Eugene M Apparatus and methods for polychronous encoding and multiplexing in neuronal prosthetic devices
US20120308076A1 (en) 2010-03-26 2012-12-06 Filip Lukasz Piekniewski Apparatus and methods for temporally proximate object recognition
US20120308136A1 (en) 2010-03-26 2012-12-06 Izhikevich Eugene M Apparatus and methods for pulse-code invariant object recognition
US8346692B2 (en) 2005-12-23 2013-01-01 Societe De Commercialisation Des Produits De La Recherche Appliquee-Socpra-Sciences Et Genie S.E.C. Spatio-temporal pattern recognition using a spiking neural network and processing thereof on a portable and/or distributed computer
US20130046716A1 (en) 2011-08-16 2013-02-21 Qualcomm Incorporated Method and apparatus for neural temporal coding, learning and recognition
US8390707B2 (en) 2008-02-28 2013-03-05 Kabushiki Kaisha Toshiba Solid-state imaging device and manufacturing method thereof
US20130073499A1 (en) 2011-09-21 2013-03-21 Eugene M. Izhikevich Apparatus and method for partial evaluation of synaptic updates based on system events
US20130073484A1 (en) 2011-09-21 2013-03-21 Eugene M. Izhikevich Elementary network description for efficient memory management in neuromorphic systems
US20130073492A1 (en) 2011-09-21 2013-03-21 Eugene M. Izhikevich Elementary network description for efficient implementation of event-triggered plasticity rules in neuromorphic systems
US20130073496A1 (en) 2011-09-21 2013-03-21 Botond Szatmary Tag-based apparatus and methods for neural networks
US20130073500A1 (en) 2011-09-21 2013-03-21 Botond Szatmary High level neuromorphic network description apparatus and methods
US20130073498A1 (en) 2011-09-21 2013-03-21 Eugene M. Izhikevich Elementary network description for efficient link between neuronal models and neuromorphic systems
US20130073491A1 (en) 2011-09-21 2013-03-21 Eugene M. Izhikevich Apparatus and methods for synaptic update in a pulse-coded network
US20130073495A1 (en) 2011-09-21 2013-03-21 Eugene M. Izhikevich Elementary network description for neuromorphic systems
US8416847B2 (en) 1998-12-21 2013-04-09 Zin Stai Pte. In, Llc Separate plane compression using plurality of compression methods including ZLN and ZLD methods
US20130117212A1 (en) 2011-11-09 2013-05-09 Qualcomm Incorporated Methods and apparatus for unsupervised neural replay, learning refinement, association and memory transfer: neural associative learning, pattern completion, separation, generalization and hierarchical replay
US20130151450A1 (en) 2011-12-07 2013-06-13 Filip Ponulak Neural network apparatus and methods for signal conversion
US20130218821A1 (en) 2011-09-21 2013-08-22 Botond Szatmary Round-trip engineering apparatus and methods for neural networks
US20130297542A1 (en) 2012-05-07 2013-11-07 Filip Piekniewski Sensory input processing apparatus in a spiking neural network
US20130297541A1 (en) 2012-05-07 2013-11-07 Filip Piekniewski Spiking neural network feedback apparatus and methods
US20130297539A1 (en) 2012-05-07 2013-11-07 Filip Piekniewski Spiking neural network object recognition apparatus and methods
US20130325774A1 (en) 2012-06-04 2013-12-05 Brain Corporation Learning stochastic apparatus and methods
US20130325775A1 (en) 2012-06-04 2013-12-05 Brain Corporation Dynamically reconfigurable stochastic learning apparatus and methods
US20130325773A1 (en) 2012-06-04 2013-12-05 Brain Corporation Stochastic apparatus and methods for implementing generalized learning rules
US20130325766A1 (en) 2012-06-04 2013-12-05 Csaba Petre Spiking neuron network apparatus and methods
US20130325768A1 (en) * 2012-06-04 2013-12-05 Brain Corporation Stochastic spiking network learning apparatus and methods
US20130325777A1 (en) 2012-06-04 2013-12-05 Csaba Petre Spiking neuron network apparatus and methods
US20140012788A1 (en) 2012-07-03 2014-01-09 Filip Piekniewski Conditional plasticity spiking neuron network apparatus and methods
US20140016858A1 (en) 2012-07-12 2014-01-16 Micah Richert Spiking neuron network sensory processing apparatus and methods
US20140032459A1 (en) 2012-07-27 2014-01-30 Brain Corporation Apparatus and methods for generalized state-dependent learning in spiking neuron networks
US20140032458A1 (en) 2012-07-27 2014-01-30 Oleg Sinyavskiy Apparatus and methods for efficient updates in spiking neuron network
US20140052679A1 (en) 2011-09-21 2014-02-20 Oleg Sinyavskiy Apparatus and methods for implementing event-based updates in spiking neuron networks
US20140064609A1 (en) 2010-08-26 2014-03-06 Csaba Petre Sensory input processing apparatus and methods
US20140122397A1 (en) 2012-10-25 2014-05-01 Brain Corporation Adaptive plasticity apparatus and methods for spiking neuron network
US20140122398A1 (en) 2012-10-25 2014-05-01 Brain Corporation Modulated plasticity apparatus and methods for spiking neuron network
US20140122399A1 (en) 2012-10-25 2014-05-01 Brain Corporation Apparatus and methods for activity-based plasticity in a spiking neuron network
US20140156574A1 (en) 2012-11-30 2014-06-05 Brain Corporation Rate stabilization through plasticity in spiking neuron network

Patent Citations (120)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5652594A (en) 1967-12-28 1997-07-29 Lockheed Martin Corporation Signal processor affording improved immunity to medium anomalies and interference in remote object detection system
US5063603A (en) 1989-11-06 1991-11-05 David Sarnoff Research Center, Inc. Dynamic method for recognizing objects and image processing system therefor
US5216752A (en) 1990-12-19 1993-06-01 Baylor College Of Medicine Interspike interval decoding neural network
US5138447A (en) 1991-02-11 1992-08-11 General Instrument Corporation Method and apparatus for communicating compressed digital video signals using multiple processors
US5272535A (en) 1991-06-13 1993-12-21 Loral Fairchild Corporation Image sensor with exposure control, selectable interlaced, pseudo interlaced or non-interlaced readout and video compression
US5875108A (en) 1991-12-23 1999-02-23 Hoffberg; Steven M. Ergonomic man-machine interface incorporating adaptive pattern recognition based control system
US6418424B1 (en) 1991-12-23 2002-07-09 Steven M. Hoffberg Ergonomic man-machine interface incorporating adaptive pattern recognition based control system
US5355435A (en) 1992-05-18 1994-10-11 New Mexico State University Technology Transfer Corp. Asynchronous temporal neural processing element
US5673367A (en) 1992-10-01 1997-09-30 Buckley; Theresa M. Method for neural network control of motion using real-time environmental feedback
US5638359A (en) 1992-12-14 1997-06-10 Nokia Telecommunications Oy Method for congestion management in a frame relay network and a node in a frame relay network
RU2108612C1 (en) 1994-09-14 1998-04-10 Круглов Сергей Петрович Adaptive control system with identifier and implicit reference model
US6625317B1 (en) 1995-09-12 2003-09-23 Art Gaffin Visual imaging system and method
US6014653A (en) 1996-01-26 2000-01-11 Thaler; Stephen L. Non-algorithmically implemented artificial neural networks and components thereof
US6009418A (en) 1996-05-02 1999-12-28 Cooper; David L. Method and apparatus for neural networking using semantic attractor architecture
US6509854B1 (en) 1997-03-16 2003-01-21 Hitachi, Ltd. DA conversion circuit
US20030050903A1 (en) 1997-06-11 2003-03-13 Jim-Shih Liaw Dynamic synapse for signal processing in neural networks
US6545708B1 (en) 1997-07-11 2003-04-08 Sony Corporation Camera controlling device and method for predicted viewing
US6458157B1 (en) 1997-08-04 2002-10-01 Suaning Gregg Joergen Retinal stimulator
US6581046B1 (en) 1997-10-10 2003-06-17 Yeda Research And Development Co. Ltd. Neuronal phase-locked loops
US6545705B1 (en) 1998-04-10 2003-04-08 Lynx System Developers, Inc. Camera with object recognition/data output
US6035389A (en) 1998-08-11 2000-03-07 Intel Corporation Scheduling instructions with different latencies
US20040170330A1 (en) 1998-08-12 2004-09-02 Pixonics, Inc. Video coding reconstruction apparatus and methods
US8416847B2 (en) 1998-12-21 2013-04-09 Zin Stai Pte. In, Llc Separate plane compression using plurality of compression methods including ZLN and ZLD methods
US6546291B2 (en) 2000-02-16 2003-04-08 Massachusetts Eye & Ear Infirmary Balance prosthesis
US7054850B2 (en) 2000-06-16 2006-05-30 Canon Kabushiki Kaisha Apparatus and method for detecting or recognizing pattern by employing a plurality of feature detecting elements
US7737933B2 (en) 2000-09-26 2010-06-15 Toshiba Matsushita Display Technology Co., Ltd. Display unit and drive system thereof and an information display unit
US20040193670A1 (en) 2001-05-21 2004-09-30 Langan John D. Spatio-temporal filter and method
US20050036649A1 (en) 2001-08-23 2005-02-17 Jun Yokono Robot apparatus, face recognition method, and face recognition apparatus
US20030216919A1 (en) 2002-05-13 2003-11-20 Roushar Joseph C. Multi-dimensional method and apparatus for automated language interpretation
US20040136439A1 (en) 2002-11-15 2004-07-15 Brandon Dewberry Methods and systems acquiring impulse signals
US20060094001A1 (en) 2002-11-29 2006-05-04 Torre Vicent E Method and device for image processing and learning with neuronal cultures
US20040220082A1 (en) * 2003-01-21 2004-11-04 Northwestern University Manipulation of neuronal ion channels
US20080199072A1 (en) 2003-02-27 2008-08-21 Sony Corporation Image processing device and method, learning device and method, recording medium, and program
US20090043722A1 (en) 2003-03-27 2009-02-12 Alex Nugent Adaptive neural network utilizing nanotechnology-based components
US20080100482A1 (en) 2003-05-27 2008-05-01 Lazar Aurel A Multichannel Time Encoding And Decoding Of A Signal
US20050015351A1 (en) 2003-07-18 2005-01-20 Alex Nugent Nanotechnology neural network methods and systems
US20060161218A1 (en) 2003-11-26 2006-07-20 Wicab, Inc. Systems and methods for treating traumatic brain injury
US7580907B1 (en) 2004-01-14 2009-08-25 Evolved Machines, Inc. Invariant object recognition
US20070208678A1 (en) 2004-03-17 2007-09-06 Canon Kabushiki Kaisha Parallel Pulse Signal Processing Apparatus, Pattern Recognition Apparatus, And Image Input Apparatus
US7653255B2 (en) 2004-06-02 2010-01-26 Adobe Systems Incorporated Image region of interest encoding
US8015130B2 (en) 2004-06-11 2011-09-06 Canon Kabushiki Kaisha Information processing apparatus, information processing method, pattern recognition apparatus, and pattern recognition method
US20050283450A1 (en) 2004-06-11 2005-12-22 Masakazu Matsugu Information processing apparatus, information processing method, pattern recognition apparatus, and pattern recognition method
US20060129728A1 (en) 2004-12-09 2006-06-15 Hampel Craig E Memory interface with workload adaptive encode/decode
US8000967B2 (en) 2005-03-09 2011-08-16 Telefonaktiebolaget Lm Ericsson (Publ) Low-complexity code excited linear prediction encoding
US20080201282A1 (en) 2005-03-31 2008-08-21 France Telecom System and Method for Locating Points of Interest in an Object Image Implementing a Neural Network
US20070176643A1 (en) 2005-06-17 2007-08-02 Alex Nugent Universal logic gate utilizing nanotechnology
US20070022068A1 (en) 2005-07-01 2007-01-25 Ralph Linsker Neural networks for prediction and control
RU2424561C2 (en) 2005-08-31 2011-07-20 Майкрософт Корпорейшн Training convolutional neural network on graphics processing units
US7765029B2 (en) 2005-09-13 2010-07-27 Neurosciences Research Foundation, Inc. Hybrid control device
US8583286B2 (en) 2005-09-13 2013-11-12 Neurosciences Research Foundation, Inc. Hybrid control device
US8154436B2 (en) 2005-10-24 2012-04-10 Mitsubishi Electric Information Technology Centre Europe B.V. Object detection
US8346692B2 (en) 2005-12-23 2013-01-01 Societe De Commercialisation Des Produits De La Recherche Appliquee-Socpra-Sciences Et Genie S.E.C. Spatio-temporal pattern recognition using a spiking neural network and processing thereof on a portable and/or distributed computer
US7849030B2 (en) 2006-05-31 2010-12-07 Hartford Fire Insurance Company Method and system for classifying documents
RU2406105C2 (en) 2006-06-13 2010-12-10 Филипп Геннадьевич Нестерук Method of processing information in neural networks
US20100081958A1 (en) 2006-10-02 2010-04-01 She Christy L Pulse-based feature extraction for neural recordings
JP4087423B2 (en) 2006-10-17 2008-05-21 京セラミタ株式会社 Portable communication device
US8103602B2 (en) 2006-12-29 2012-01-24 Neurosciences Research Foundation, Inc. Solving the distal reward problem through linkage of STDP and dopamine signaling
WO2008083335A2 (en) 2006-12-29 2008-07-10 Neurosciences Research Foundation, Inc. Solving the distal reward problem through linkage of stdp and dopamine signaling
US20100100482A1 (en) 2007-01-23 2010-04-22 Sxip Identity Corp. Intermediate Data Generation For Transaction Processing
US20080237446A1 (en) 2007-02-16 2008-10-02 Texas Instruments Incorporated Solid-state image pickup device and method
WO2008132066A1 (en) 2007-04-27 2008-11-06 Siemens Aktiengesellschaft A method for computer-assisted learning of one or more neural networks
US20100225824A1 (en) 2007-06-28 2010-09-09 The Trustees Of Columbia University In The City Of New York Multi-Input Multi-Output Time Encoding And Decoding Machines
US8281997B2 (en) 2008-02-19 2012-10-09 Bilcare Technologies Singapore Pte. Ltd. Reading device for identifying a tag or an object adapted to be identified, related methods and systems
US8390707B2 (en) 2008-02-28 2013-03-05 Kabushiki Kaisha Toshiba Solid-state imaging device and manufacturing method thereof
US20110160741A1 (en) 2008-06-09 2011-06-30 Hiroyuki Asano Medical treatment tool for tubular organ
US20100036457A1 (en) 2008-08-07 2010-02-11 Massachusetts Institute Of Technology Coding for visual prostheses
US20110137843A1 (en) 2008-08-28 2011-06-09 Massachusetts Institute Of Technology Circuits and Methods Representative of Spike Timing Dependent Plasticity of Neurons
US20100086171A1 (en) 2008-10-02 2010-04-08 Silverbrook Research Pty Ltd Method of imaging coding pattern having merged data symbols
US20100166320A1 (en) 2008-12-26 2010-07-01 Paquier Williams J F Multi-stage image pattern recognizer
US8160354B2 (en) 2008-12-26 2012-04-17 Five Apes, Inc. Multi-stage image pattern recognizer
US20100235310A1 (en) 2009-01-27 2010-09-16 Gage Fred H Temporally dynamic artificial neural networks
US20100299296A1 (en) 2009-05-21 2010-11-25 International Business Machines Corporation Electronic learning synapse with spike-timing dependent plasticity using unipolar memory-switching elements
US20110016071A1 (en) 2009-07-20 2011-01-20 Guillen Marcos E Method for efficiently simulating the information processing in cells and tissues of the nervous system with a temporal series compressed encoding neural network
US8200593B2 (en) 2009-07-20 2012-06-12 Corticaldb Inc Method for efficiently simulating the information processing in cells and tissues of the nervous system with a temporal series compressed encoding neural network
US20110119215A1 (en) 2009-11-13 2011-05-19 International Business Machines Corporation Hardware analog-digital neural networks
US8311965B2 (en) 2009-11-18 2012-11-13 International Business Machines Corporation Area efficient neuromorphic circuits using field effect transistors (FET) and variable resistance material
US20110119214A1 (en) 2009-11-18 2011-05-19 International Business Machines Corporation Area efficient neuromorphic circuits
US20120308136A1 (en) 2010-03-26 2012-12-06 Izhikevich Eugene M Apparatus and methods for pulse-code invariant object recognition
US8315305B2 (en) 2010-03-26 2012-11-20 Brain Corporation Systems and methods for invariant pulse latency coding
US20120303091A1 (en) 2010-03-26 2012-11-29 Izhikevich Eugene M Apparatus and methods for polychronous encoding and multiplexing in neuronal prosthetic devices
US20120308076A1 (en) 2010-03-26 2012-12-06 Filip Lukasz Piekniewski Apparatus and methods for temporally proximate object recognition
US20130251278A1 (en) 2010-03-26 2013-09-26 Eugene M. Izhikevich Invariant pulse latency coding systems and methods
US8467623B2 (en) 2010-03-26 2013-06-18 Brain Corporation Invariant pulse latency coding systems and methods systems and methods
US20120011090A1 (en) 2010-07-07 2012-01-12 Qualcomm Incorporated Methods and systems for three-memristor synapse with stdp and dopamine signaling
US20140064609A1 (en) 2010-08-26 2014-03-06 Csaba Petre Sensory input processing apparatus and methods
US20120084240A1 (en) 2010-09-30 2012-04-05 International Business Machines Corporation Phase change memory synaptronic circuit for spiking computation, association and recall
US20120083982A1 (en) 2010-10-05 2012-04-05 Zachary Thomas Bonefas System and method for governing a speed of an autonomous vehicle
US20120109866A1 (en) 2010-10-29 2012-05-03 International Business Machines Corporation Compact cognitive synaptic computing circuits
CN102226740A (en) 2011-04-18 2011-10-26 中国计量学院 Bearing fault detection method based on manner of controlling stochastic resonance by external periodic signal
US20130046716A1 (en) 2011-08-16 2013-02-21 Qualcomm Incorporated Method and apparatus for neural temporal coding, learning and recognition
US20130218821A1 (en) 2011-09-21 2013-08-22 Botond Szatmary Round-trip engineering apparatus and methods for neural networks
US20130073484A1 (en) 2011-09-21 2013-03-21 Eugene M. Izhikevich Elementary network description for efficient memory management in neuromorphic systems
US20130073495A1 (en) 2011-09-21 2013-03-21 Eugene M. Izhikevich Elementary network description for neuromorphic systems
US20130073498A1 (en) 2011-09-21 2013-03-21 Eugene M. Izhikevich Elementary network description for efficient link between neuronal models and neuromorphic systems
US20140052679A1 (en) 2011-09-21 2014-02-20 Oleg Sinyavskiy Apparatus and methods for implementing event-based updates in spiking neuron networks
US20130073491A1 (en) 2011-09-21 2013-03-21 Eugene M. Izhikevich Apparatus and methods for synaptic update in a pulse-coded network
US20130073500A1 (en) 2011-09-21 2013-03-21 Botond Szatmary High level neuromorphic network description apparatus and methods
US20130073496A1 (en) 2011-09-21 2013-03-21 Botond Szatmary Tag-based apparatus and methods for neural networks
US20130073492A1 (en) 2011-09-21 2013-03-21 Eugene M. Izhikevich Elementary network description for efficient implementation of event-triggered plasticity rules in neuromorphic systems
US20130073499A1 (en) 2011-09-21 2013-03-21 Eugene M. Izhikevich Apparatus and method for partial evaluation of synaptic updates based on system events
US8712941B2 (en) 2011-09-21 2014-04-29 Brain Corporation Elementary network description for efficient link between neuronal models and neuromorphic systems
US20130117212A1 (en) 2011-11-09 2013-05-09 Qualcomm Incorporated Methods and apparatus for unsupervised neural replay, learning refinement, association and memory transfer: neural associative learning, pattern completion, separation, generalization and hierarchical replay
US20130151450A1 (en) 2011-12-07 2013-06-13 Filip Ponulak Neural network apparatus and methods for signal conversion
US20130297539A1 (en) 2012-05-07 2013-11-07 Filip Piekniewski Spiking neural network object recognition apparatus and methods
US20130297541A1 (en) 2012-05-07 2013-11-07 Filip Piekniewski Spiking neural network feedback apparatus and methods
US20130297542A1 (en) 2012-05-07 2013-11-07 Filip Piekniewski Sensory input processing apparatus in a spiking neural network
US20130325773A1 (en) 2012-06-04 2013-12-05 Brain Corporation Stochastic apparatus and methods for implementing generalized learning rules
US20130325766A1 (en) 2012-06-04 2013-12-05 Csaba Petre Spiking neuron network apparatus and methods
US20130325777A1 (en) 2012-06-04 2013-12-05 Csaba Petre Spiking neuron network apparatus and methods
US20130325768A1 (en) * 2012-06-04 2013-12-05 Brain Corporation Stochastic spiking network learning apparatus and methods
US20130325774A1 (en) 2012-06-04 2013-12-05 Brain Corporation Learning stochastic apparatus and methods
US20130325775A1 (en) 2012-06-04 2013-12-05 Brain Corporation Dynamically reconfigurable stochastic learning apparatus and methods
US20140012788A1 (en) 2012-07-03 2014-01-09 Filip Piekniewski Conditional plasticity spiking neuron network apparatus and methods
US20140016858A1 (en) 2012-07-12 2014-01-16 Micah Richert Spiking neuron network sensory processing apparatus and methods
US20140032458A1 (en) 2012-07-27 2014-01-30 Oleg Sinyavskiy Apparatus and methods for efficient updates in spiking neuron network
US20140032459A1 (en) 2012-07-27 2014-01-30 Brain Corporation Apparatus and methods for generalized state-dependent learning in spiking neuron networks
US20140122397A1 (en) 2012-10-25 2014-05-01 Brain Corporation Adaptive plasticity apparatus and methods for spiking neuron network
US20140122398A1 (en) 2012-10-25 2014-05-01 Brain Corporation Modulated plasticity apparatus and methods for spiking neuron network
US20140122399A1 (en) 2012-10-25 2014-05-01 Brain Corporation Apparatus and methods for activity-based plasticity in a spiking neuron network
US20140156574A1 (en) 2012-11-30 2014-06-05 Brain Corporation Rate stabilization through plasticity in spiking neuron network

Non-Patent Citations (96)

* Cited by examiner, † Cited by third party
Title
Berkes and Wiskott, Slow feature analysis yields a rich repertoire of complex cell properties. Journal of Vision (2005) vol. 5 (6).
Bohte, 'Spiking Nueral Networks' Doctorate at the University of Leiden, Holland, Mar. 5, 2003, pp. 1-133 [retrieved on Nov. 14, 2012]. Retrieved from the internet: .
Bohte, 'Spiking Nueral Networks' Doctorate at the University of Leiden, Holland, Mar. 5, 2003, pp. 1-133 [retrieved on Nov. 14, 2012]. Retrieved from the internet: <URL: http://holnepagcs,cwi ,n11-sbolltedmblica6ond)hdthesislxif>.
Brette et al., Brian: a simple and flexible simulator for spiking neural networks, The Neuromorphic Engineer, Jul. 1,2009, pp. 1-4, doi: 10.2417/1200906.1659.
Brette, et al., "Simulation of Networks of Spiking Neurons: A Review of Tools and Strategies", Received Nov. 29, 2006, Revised Apr. 2, 2007, Accepted Apr. 12, 2007, Springer Science, 50 pages.
Cessac et al. 'Overview of facts and issues about neural coding by spikes.' Journal of Physiology, Paris 104.1 (2010): 5.
Chistiakova, Marina, and Maxim Volgushev. "Heterosynaptic plasticity in the neocortex." Experimental brain research 199.3-4 (2009): 377-390.
Cuntz et al., 'One Rule to Grow Them All: A General Theory of Neuronal Branching and Its Paractical Application' PLOS Computational Biology, 6 (8), Published Aug. 5, 2010.
Davison et al., PyNN: a common interface for neuronal network simulators. Frontiers in Neuroinformatics, Jan. 2009, pp. 1-10, vol. 2, Article 11.
Djurfeldt, Mikael, The Connection-set Algebra: a formalism for the representation of connectivity structure in neuronal network models, implementations in Python and C++, and their use in simulators BMC Neuroscience Jul. 18, 2011 p. 1 12(Suppl 1):P80.
Dorval et al. 'Probability distributions of the logarithm of inter-spike intervals yield accurate entropy estimates from small datasets.' Journal of neuroscience methods 173.1 (2008): 129.
Fidjeland et al., 'Accelerated Simulation of Spiking Neural Networks Using GPUs [online],2010 [retrieved on Jun. 15, 2013], Retrieved from the Internet: URL:http://ieeexplore.ieee.org/xpls/abs-all.jsp?ammber=5596678&tag=1.
Field, G.; Chichilnisky, E., Information Processing in the Primate Retina: Circuitry and Coding. Annual Review of Neuroscience, 2007, 30(1), 1-30.
Fiete, et al. "Spike-time-dependent plasticity and heterosynaptic competition organize networks to produce long scale-free sequences of neural activity." Neuron 65.4 (2010): 563-576.
Fiete, et al., Spike-Time-Dependent Plasticity and Heterosynaptic Competition Organize Networks to Produce Long Scale-Free Sequences of Neural Activity. Neuron 65, Feb. 25, 2010, pp. 563-576.
Floreano et al., 'Neuroevolution: from architectures to learning' Evol. Intel. Jan 2008 1:47-62, [retrieved Dec. 30, 2013] [retrieved online from URL:.
Floreano et al., 'Neuroevolution: from architectures to learning' Evol. Intel. Jan 2008 1:47-62, [retrieved Dec. 30, 2013] [retrieved online from URL:<http://inforscience.epfl.ch/record/112676/files/FloreanoDuerrMattiussi2008.p df>.
Földiak, P., Learning invariance from transformation sequences. Neural Computation, 1991, 3(2), 194-200.
Froemke et al., 'Temporal modulation of spike-timing-dependent plasticity, Frontiers in Synaptic Neuroscience, vol. 2, Article 19, pp. 1-16 [online] Jun. 2010 [retrieved on Dec. 16, 2013]. Retrieved from the internet: .
Froemke et al., 'Temporal modulation of spike-timing-dependent plasticity, Frontiers in Synaptic Neuroscience, vol. 2, Article 19, pp. 1-16 [online] Jun. 2010 [retrieved on Dec. 16, 2013]. Retrieved from the internet: <frontiersin.org>.
Gerstner et al. (1996) A neuronal learning rule for sub-millisecond temporal coding. Nature vol. 383 (6595) pp. 76-78.
Gewaltig et al., 'NEST (Neural Simulation Tool)', Scholarpedia, 2007, pp. 1-15, 2(4): 1430, dol: 1 0.4249/scholarpedia.1430.
Gleeson et al., NeuroML: A Language for Describing Data Driven Models of Neurons and Networks with a High Degree of Biological Detail, PLoS Computational Biology, Jun. 2010, pp. 1-19 vol. 6 Issue 6.
Gluck, Stimulus Generalization and Representation in Adaptive Network Models of Category Learning [online], 1991 [retrieved on Aug. 24, 2013]. Retrieved from the lnternet:<URL:http:// www.google.com/url?sa=t&rct=j&q=Giuck+%22STIMULUS+GENERALIZATION+AND+REPRESENTATIO N+1 N +ADAPTIVE+NETWORK+MODELS+OF+CATEGORY+LEARN I NG%22+ 1991.
Gollisch et al. 'Rapid neural coding in the retina with relative spike latencies.' Science 319.5866 (2008): 11 08-1111.
Goodman et al., Brian: a simulator for spiking neural networks in Python, Frontiers in Neuroinformatics, Nov. 2008, pp. 1-10, vol. 2, Article 5.
Gorchetchnikov et al., NineML: declarative, mathematically-explicit descriptions of spiking neuronal networks, Frontiers in Neuroinformatics, Conference Abstract: 4th INCF Congress of Neuroinformatics, doi: 1 0.3389/conf fninf.2011.08.00098.
Graham, Lyle J., The Surf-Hippo Reference Manual, http:// www.neurophys.biomedicale.univparis5. fr/-graham/surf-hippo-files/Surf-Hippo%20Reference%20Manual.pdf, Mar. 2002, pp. 1-128.
Hopfield JJ (1995) Pattern recognition computation using action potential timing for stimulus representation. Nature 376: 33-36.
Itti, Laurent, and Christof Koch. "Computational modelling of visual attention." Nature reviews neuroscience 2.3 (2001): 194-203.
Izhikevich E. M. and Hoppensteadt F.C. (2009) Polychronous Wavefront Computations. International Journal of Bifurcation and Chaos, 19:1733-1739.
Izhikevich E.M. (2004) Which Model to Use for Cortical Spiking Neurons? IEEE Transactions on Neural Networks, 15:1063-1070.
Izhikevich E.M. (2006) Polychronization: Computation With Spikes. Neural Computation,18:245-282.
Izhikevich E.M., "Neural Excitability, Spiking and Bursting", Neurosciences Institute, Received Jun. 9, 1999, Revised Oct. 25, 1999, 1171-1266, 96 pages.
Izhikevich et al., 'Relating STDP to BCM', Neural Computation (2003) 15, 1511-1523.
Izhikevich, E.M. (2007) Dynamical Systems in Neuroscience: The Geometry of Excitability and Bursting, The MIT Press, 2007.
Izhikevich, Eugene M. "Simple model of spiking neurons." IEEE Transactions on neural networks. 14.6 (2003): 1569-1572.
Izhikevich, Eugene M. Dynamical systems in neuroscience: chapters 1 and 2. MIT press. 2007.
Izhikevich, 'Simple Model of Spiking Neurons', IEEE Transactions on Neural Networks, vol. 14, No. 6, Nov. 2003, pp. 1569-1572.
Janowitz, M.K.; Van Rossum, M.C.W. Excitability changes that complement Hebbian learning. Network, Computation in Neural Systems, 2006, 17 (1). 31-41.
Karbowski et al., 'Multispikes and Synchronization in a Large Neural Network with Temporal Delays', Neural Computation 12, 1573-1606 (2000).
Kazantsev, et al., "Active Spike Transmission in the Neuron Model With a Winding Threshold Maniford", Jan. 3, 2012, 205-211, 7 pages.
Khotanzad, Alireza, Classification of invariant image representations using a neural network, IEEF. Transactions on Acoustics, Speech, and Signal Processing vol. 38 No. 6 Jun. 1990 pp. 1028-1038.
Khotanzad, 'Classification of invariant image representations using a neural network' IEEF. Transactions on Acoustics, Speech, and Signal Processing, vol. 38, No. 6, Jun. 1990, pp. 1028-1038 [online], [retrieved on Dec. 10, 2013]. Retrieved from the Internet .
Khotanzad, 'Classification of invariant image representations using a neural network' IEEF. Transactions on Acoustics, Speech, and Signal Processing, vol. 38, No. 6, Jun. 1990, pp. 1028-1038 [online], [retrieved on Dec. 10, 2013]. Retrieved from the Internet <URL: http://www-ee.uta.edu/eeweb/IP/Courses/SPR/Reference/ Khotanzad.pdf>.
Kling-Petersen, PhD, "Sun and HPC: From Systems to PetaScale" Sun Microsystems, no date, 31 pages.
Knoblauch, Andreas, Gunther Palm, and Friedrich T. Sommer. "Memory capacities for synaptic and structural plasticity." Neural Computation 22.2 (2010): 289-341.
Knoblauch, et al. Memory Capacities for Synaptic and Structural Plasticity, Neural Computation 2009, pp. 1-45.
Laurent, 'Issue I-nnql-Refactor Nucleus into its own file-Neural Network Query Language' [retrieved on Nov. 12, 2013]. Retrieved from the Internet: URL:https:// code.google.com/p/nnql/issues/detail?id=1.
Laurent, 'The Neural Network Query Language (NNQL) Reference' [retrieved on Nov. 12, 2013]. Retrieved from the Internet: .
Laurent, 'The Neural Network Query Language (NNQL) Reference' [retrieved on Nov. 12, 2013]. Retrieved from the Internet: <URL'https://code.google.com/p/nnql/issues/detail?id=1>.
Lazar et al. 'A video time encoding machine', in Proceedings of the 15th IEEE International Conference on Image Processing (ICIP '08), 2008, pp. 717-720.
Lazar et al. 'Multichannel time encoding with integrate-and-fire neurons.' Neurocomputing 65 (2005): 401-407.
Lazar et al.,'Consistent recovery of sensory stimuli encoded with MIMO neural circuits.' Computational intelligence and neuroscience (2010): 2.
Li, Zhaoping. "A saliency map in primary visual cortex." Trends in cognitive sciences 6.1 (2002): 9-16.
Markram, Henry, et al. "Regulation of synaptic efficacy by coincidence of postsynaptic APs and EPSPs." Science 275.5297 (1997): 213-215.
Martinez-Perez, et al., "Automatic Activity Estimation Based on Object Behavior Signature", 2010, 10 pages.
Masquelier and Thorpe, Learning to recognize objects using waves of spikes and Spike Timing-Dependent Plasticity. Neural Networks (IJCNN), The 2010 International Joint Conference on DOI -10.1109/IJCNN.2010.5596934 (2010) pp. 1-8.
Masquelier, Timothee. 'Relative spike time coding and STOP-based orientation selectivity in the early visual system in natural continuous and saccadic vision: a computational model.' Journal of computational neuroscience 32.3 (2012): 425-441.
Matsugu, et al., "Convolutional Spiking Neural Network for Robust Object Detection with Population Code Using Structured Pulse Packets", 2004, 39-55, 17 pages.
Meinhardt, Hans, and Alfred Gierer. "Pattern formation by local self-activation and lateral inhibition." Bioessays 22.8 (2000): 753-760.
Meister, M., Multineuronal codes in retinal signaling. Proceedings of the National Academy of sciences. 1996, 93, 609-614.
Meister, M.; Berry, M.J. The neural code of the retina, Neuron. 1999, 22, 435-450.
Nichols, A Re configurable Computing Architecture for Implementing Artificial Neural Networks on FPGA, Master's Thesis, The University of Guelph, 2003, pp. 1-235.
Oster M., Lichtsteiner P., Delbruck T, Liu S. A Spike-Based Saccadic Recognition System. ISCAS 2007. IEEE International Symposium on Circuits and Systems, 2009, pp. 3083-3086.
Paugam-Moisy et al. 'Computing with spiking neuron networks.' Handbook of Natural Computing, 40 p. Springer, Heidelberg (2009).
Paugam-Moisy et al., "Computing with spiking neuron networks" G. Rozenberg T. Back, J. Kok (Eds.), Handbook of Natural Computing, Springer-Verlag (2010) [retrieved Dec. 30, 2013], [retrieved online from link.springer.com].
Pavlidis et al. 'Spiking neural network training using evolutionary algorithms. In: Proceedings 2005 IEEE International Joint Conference on Neural Networkds, 2005. IJCNN'05, vol. 4, pp. 2190-2194 Publication Date Jul. 31, 2005 [online] [Retrieved on Dec. 10, 2013] Retrieved from the Internet <URL: http://citeseerx.ist.psu.edu/ viewdoc/download?doi=10.1.1.5.4346&rep=rep1&type=pdf.
Ramachandran, et al., "The Perception of Phantom Limbs", The D.O. Hebb Lecture, Center for Brain and Cognition, University of California, 1998, 121, 1603-1630, 28 pages.
Rekeczky, et al., "Cellular Multiadaptive Analogic Architecture: A Computational Framework for UAV Applications." May 2004.
Revow M., Williams C., and Hinton, G.E., 1996, Using Generative Models for Handwritten Digit Recognition, IEEE Trans. on Pattern Analysis and Machine Intelligence, 18, No. 6, Jun. 1996.
Sanchez, Efficient Simulation Scheme for Spiking Neural Networks, Doctoral Thesis, Universda di Granada Mar. 28, 2008, pp. 1-104.
Sato et al., 'Pulse interval and width modulation for video transmission.' Cable Television, IEEE Transactions on 4 (1978): 165-173.
Schemmel, J., et al. Implementing synaptic plasticity in a VLSI spiking neural network model, Proceedings of the 2006 International Joint Conference on Neural Networks, Jul. 2006 pp. 1-6.
Schnitzer, We; Meister, M.; Multineuronal Firing Patterns in the Signal from Eye to Brain. Neuron, 2003, 37, 499-511.
Serrano-Gotarredona, et al, "On Real-Time: AER 2-D Convolutions Hardware for Neuromorphic Spike-based Cortical Processing", Jul. 2008.
SIMULINK.RTM. model [online], [Retrieved on Dec. 10, 2013] Retrieved from &It:URL: http://www.mathworks.com/ products/simulink/index.html&gt.
Sinyavskiy et al. 'Reinforcement learning of a spiking neural network in the task of control of an agent in a virtual discrete environment' Rus. J. Nonlin. Dyn., 2011, vol. 7, No. 4 (Mobile Robots), pp. 859-875, chapters 1-8 (Russian Article with English Abstract).
Sjostrom et al., 'Spike-Timing Dependent Plasticity' Scholarpedia, 5(2):1362 (2010), pp. 1-18.
Stringer, et al., "Invariant Object Recognition in the Visual System with Novel Views of 3D Objects", 2002. 2585-2596, 12 pages.
Swiercz, Waldemar, et al. "A new synaptic plasticity rule for networks of spiking neurons." Neural Networks, IEEE Transactions on 17.1 (2006) 94-105.
Szatmary et al., 'Spike-timing Theory of Working Memory' PLoS Computational Biology, vol. 6, Issue 8, Aug. 19, 2010 [retrieved on Dec. 30, 2013]. Retrieved from the Internet: .
Szatmary et al., 'Spike-timing Theory of Working Memory' PLoS Computational Biology, vol. 6, Issue 8, Aug. 19, 2010 [retrieved on Dec. 30, 2013]. Retrieved from the Internet: <URL: http://www.ploscompbiol.org/article/info%3Adoi%2F10.1371 %2Fjournal.pcbi.10008 79#>.
Thomas S. and Riesenhuber, M, 2004, Realistic Modeling of Simple and Complex Cell Tuning in the HMAX Model, and Implications for Invariant Object Recognition in Cortex, Al Memo 2004-017 Jul. 2004.
Thorpe, S.J., Delorme, A. & Vanrullen, R. (2001). Spike-based strategies for rapid processing. Neural Networks 14, pp. 715-725.
Thorpe, S.J., Guyonneau, R., Guilbaud, N., Allegraud, J-M. & Vanrullen, R. (2004). SpikeNet: real-time visual processing with one spike per neuron. Neurocomputing, 58-60, pp. 857-864.
Tim Gollisch and Markus Meister (2008) Rapid Neural Coding in the Retina with Relative Spike Latencies. Science 319:1 108-1111.
Van Rullen R.; Thorpe, S. Rate Coding versus temporal order coding: What the Retinal ganglion cells tell the visual cortex. Neural computation, 2001, 13, 1255-1283.
Vanrullen, R. & Koch, C. (2003). Is perception discrete or continuous? Trends in Cognitive Sciences 7(5), pp. 207-213.
Vanrullen, R., Guyonneau, R. & Thorpe, S.J. (2005). Spike times make sense. Trends in Neurosciences 28(1).
Wallis, G.; Rolls, E. T. A model of invariant object recognition in the visual system. Progress in Neurobiology. 1997, 51, 167-194.
Wang, 'The time dimension for scene analysis.' Neural Networks, IEEE Transactions on 16.6 (2005): 1401-1426.
Wiskott, et al., "Slow Feature Analysis", 2002, 29 pages.
Wiskott, L.; Sejnowski, T.J. Slow feature analysis: Unsupervised learning of invariances. Neural Computation, 2002, 14, (4), 715-770.
Wu, QingXiang, et al. "Remembering Key Features of Visual Images based on Spike Timing Dependent Plasticity of Spiking Neurons." Image and Signal Processing, 2009. CISP '09. 2nd International Congress on, IEEE. 2009.
Zarandy, et al., "Bi-i: A Standalone Ultra High Speed Cellular Vision System", Jun. 2005.

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10387741B2 (en) 2016-12-21 2019-08-20 Volkswagen Ag Digital neuromorphic (NM) sensor array, detector, engine and methodologies
US20180173983A1 (en) * 2016-12-21 2018-06-21 Volkswagen Ag Digital neuromorphic (nm) sensor array, detector, engine and methodologies
US10133944B2 (en) * 2016-12-21 2018-11-20 Volkswagen Ag Digital neuromorphic (NM) sensor array, detector, engine and methodologies
US10229341B2 (en) * 2016-12-21 2019-03-12 Volkswagen Ag Vector engine and methodologies using digital neuromorphic (NM) data
US10235565B2 (en) 2016-12-21 2019-03-19 Volkswagen Ag System and methodologies for occupant monitoring utilizing digital neuromorphic (NM) data and fovea tracking
US10282615B2 (en) 2016-12-21 2019-05-07 Volkswagen Ag System and method for root association in image data
US20180173992A1 (en) * 2016-12-21 2018-06-21 Volkswagen Ag Vector engine and methodologies using digital neuromorphic (nm) data
US10387725B2 (en) 2016-12-21 2019-08-20 Volkswagen Ag System and methodologies for occupant monitoring utilizing digital neuromorphic (NM) data and fovea tracking
US10423856B2 (en) 2016-12-21 2019-09-24 Volkswagen Ag Vector engine and methodologies using digital neuromorphic (NM) data
US10789495B2 (en) 2016-12-21 2020-09-29 Volkswagen Ag System and method for 1D root association providing sparsity guarantee in image data
US11100396B2 (en) 2017-08-24 2021-08-24 International Business Machines Corporation Self-adjusting threshold for synaptic activity in neural networks
US11556343B2 (en) 2017-09-22 2023-01-17 International Business Machines Corporation Computational method for temporal pooling and correlation
US11080592B2 (en) 2017-10-05 2021-08-03 International Business Machines Corporation Neuromorphic architecture for feature learning using a spiking neural network
US10922824B1 (en) 2019-08-13 2021-02-16 Volkswagen Ag Object tracking using contour filters and scalers

Also Published As

Publication number Publication date
US20140122397A1 (en) 2014-05-01

Similar Documents

Publication Publication Date Title
US9183493B2 (en) Adaptive plasticity apparatus and methods for spiking neuron network
US9111226B2 (en) Modulated plasticity apparatus and methods for spiking neuron network
US9460385B2 (en) Apparatus and methods for rate-modulated plasticity in a neuron network
US9111215B2 (en) Conditional plasticity spiking neuron network apparatus and methods
US9275326B2 (en) Rate stabilization through plasticity in spiking neuron network
US9224090B2 (en) Sensory input processing apparatus in a spiking neural network
US9129221B2 (en) Spiking neural network feedback apparatus and methods
US8972315B2 (en) Apparatus and methods for activity-based plasticity in a spiking neuron network
US9098811B2 (en) Spiking neuron network apparatus and methods
US9177245B2 (en) Spiking network apparatus and method with bimodal spike-timing dependent plasticity
US9218563B2 (en) Spiking neuron sensory processing apparatus and methods for saliency detection
US9256215B2 (en) Apparatus and methods for generalized state-dependent learning in spiking neuron networks
US20130297539A1 (en) Spiking neural network object recognition apparatus and methods
US8990133B1 (en) Apparatus and methods for state-dependent learning in spiking neuron networks
US20130325766A1 (en) Spiking neuron network apparatus and methods
US9256823B2 (en) Apparatus and methods for efficient updates in spiking neuron network
US9193075B1 (en) Apparatus and methods for object detection via optical flow cancellation
US9070039B2 (en) Temporal winner takes all spiking neuron network sensory processing apparatus and methods
US9122994B2 (en) Apparatus and methods for temporally proximate object recognition
US9489623B1 (en) Apparatus and methods for backward propagation of errors in a spiking neuron network
US8977582B2 (en) Spiking neuron network sensory processing apparatus and methods
US20150074026A1 (en) Apparatus and methods for event-based plasticity in spiking neuron networks
US20140193066A1 (en) Contrast enhancement spiking neuron network sensory processing apparatus and methods
US20120308136A1 (en) Apparatus and methods for pulse-code invariant object recognition
US9552546B1 (en) Apparatus and methods for efficacy balancing in a spiking neuron network

Legal Events

Date Code Title Description
AS Assignment

Owner name: BRAIN CORPORATION, CALIFORNIA

Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:RICHERT, MICAH;SZATMARY, BOTOND;SIGNING DATES FROM 20130216 TO 20130221;REEL/FRAME:030590/0554

STCF Information on status: patent grant

Free format text: PATENTED CASE

MAFP Maintenance fee payment

Free format text: PAYMENT OF MAINTENANCE FEE, 4TH YEAR, LARGE ENTITY (ORIGINAL EVENT CODE: M1551); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY

Year of fee payment: 4

AS Assignment

Owner name: HERCULES CAPITAL, INC., CALIFORNIA

Free format text: SECURITY INTEREST;ASSIGNOR:BRAIN CORPORATION;REEL/FRAME:057851/0574

Effective date: 20211004

FEPP Fee payment procedure

Free format text: MAINTENANCE FEE REMINDER MAILED (ORIGINAL EVENT CODE: REM.); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY

FEPP Fee payment procedure

Free format text: 7.5 YR SURCHARGE - LATE PMT W/IN 6 MO, LARGE ENTITY (ORIGINAL EVENT CODE: M1555); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY

MAFP Maintenance fee payment

Free format text: PAYMENT OF MAINTENANCE FEE, 8TH YEAR, LARGE ENTITY (ORIGINAL EVENT CODE: M1552); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY

Year of fee payment: 8